University of Warwick institutional repository: http://go.warwick.ac.uk/wrap
A Thesis Submitted for the Degree of PhD at the University of Warwick
http://go.warwick.ac.uk/wrap/3505
This thesis is made available online and is protected by original copyright. Please scroll down to view the document itself. Please refer to the repository record for this item for information to help you to cite it. Our policy information is available from the repository home page. MARKET CONCENTRATION, CREDIT INSTITUTIONS AND THE MACROECONOMY
MARCO MAZZOLI Dept of Economics University of Warwick
ý- ; r' ý.
Thesis submitted to the University of Warwick for the qualification of Ph. D. in Economics
Date of submission: May 1994
Research conducted at the Department of Economics of the University of Warwick Best Copy Available
Print bound close to the spine
ý d' ., SUMMARY Chapter one is a brief discussion of a few methodological premises. The second chapter is meant to show (by means of a theoretical analysis) the effective macroeconomic relevance of oligopsony in the market for credit. This is done by using two models. In the first (simplified) model - where the behaviour of the supply function of bank credit to industrial firms is captured by a "Cobb-Douglas" reduced form - an exogenous decrease in the market power of the industrial firms on the credit market increases the effectiveness of monetary policy. In the second model, where the banking sector behaves consistently with the portfolio allocation theory, the results are weakened: it is still true that, apart from extreme cases, reductions in the market power of industrial firms in the credit markets increase the macroeconomic level of investment and affect the monetary policy multiplier, but the sign of the latter effect becomes ambiguous and depends on the analytical forms of the behavioural functions. Both models, however, show that modifications of the market structure in the banking sector have, in general, macroeconomic effects. The third chapter suggests an interpretation of the phenomenon of "securitization" on the basis of Williamson's [1985] contractual framework. It is pointed out that in securitized financial systems substitutability between securities and intermediated credit is an empirically relevant phenomenon that makes the demand for bank credit to industry more unstable than the supply. For this purpose, a comparative econometric analysis has been performed with British and German data, because the two countries had (apart from the phenomenon of securitization) many similarities in their regulatory systems, as well as in the degree of concentration of their banking sectors and in the magnitude of the respective economies, at least until German Unification. The analytical form of the bank credit supply function is based on the "credit view". This specific aspect of the behaviour of banks is analyzed in Chapter 4, which contains an empirical analysis (performed with Italian data) of the free liquidity ratio for commercial banks, interpreted on the basis of the recent literature on investment decisions under conditions of investments' irreversibility and uncertainty. Chapters 5 and 6 examine the interactions between industrial firms and financial intermediaries in a "microeconomic" perspective. The focus is on the investment decision, and one of the main concerns is to perform a theoretical and empirical analysis on the connections between risk, cost of capital and investment decisions. Chapter 5 contains an empirical analysis of the firms' investment decision based on a theoretical model where the decisions concerning investment and the firms' financial structure are taken simultaneously. The results are not conclusive, in part because of the complexity of the causal links among market structure, investment and financing decisions suggested by various contributions in finance as well as in industrial economics. The study of such causal links is precisely the concern of Chapter 6, which contains an analysis of the implications of a few alternative hypotheses (based on precise results of the industrial economics literature) on the link existing between the cost of capital, the market structure and the profit margins. TABLE OF CONTENTS
l CHAPTER 1- INTRODUCTION ...... p. the 2 1. The purpose of analysis ...... 2 2. A few methodological premises ...... 8 3. Starting points and general assumptions ...... the 17 4. Structure of work ...... Chapter 1 24 Bibliography of ......
CHAPTER 2- MONETARY POLICY WITH OLIGOPSONY IN THE MARKET FOR
26 CREDIT ......
1. Introduction 27 ......
2. Monetary policy in a simplified model with oligopsonistic industrial firms in the 29 credit market ......
3. banking households 34 A generalization: sector and ...... Monetary Authority 35 3.1'The ...... 37 3.2 Households ...... Sector 39 3.3 The Banking ......
Industrial Sector 45 3.4 The ...... Conditions 47 3.5 Equilibrium ...... Statics 49 3.6 Comparative ...... 56 4. Conclusions ...... Chapter 2 58 Bibliography of ...... 59 APPENDIX ......
CHAPTER 3- CAPITAL MARKETS' SOPHISTICATION AND BANK LENDING: AN
INTERPRETATION IN THE SPIRIT OF 0. WILLIAMSON (1985) AND AN
63 EMPIRICAL ANALYSIS ...... 64 1. Introduction...... 2 An interpretation of securitization based on 0. Williamson's
framework 66 [1985] contractual relations ......
2.1 Securitization: empirical relevance for bank credit and some
implications 71 macroeconomic ......
3. An empirical analysis of some macroeconomic implications of
the United Kingdom Germany 76 securitization: and ......
3.1 A Brief description of the econometric methodology...... 82
83 3.2 General-to-specific and partial adjustment ...... 84 4. The Empirical Results ...... 4.1 The United Kingdom 85 ...... Germany 89 4.2 ...... 4.3 the demand for 92 A quick comparison with money ...... 5. Conclusions 95 ...... Chapter 3 99 Bibliography of ...... 102 APPENDIX ......
CHAPTER 4- IS IT MONEY, CREDIT OR BOTH, OR NEITHER? AN EMPIRICAL
INVESTIGATION BASED ON THE FREE LIQUIDITY RATIO FOR COMMERCIAL
104 BANKS ...... 105 1. Introduction ...... few 107 2. A comments on some standard empirical works ......
3. A new possible interpretation of the free liquidity ratio for
banks 110 commercial ...... 115 4. The model ......
5. 128 Conclusions ...... Chapter 4 130 Bibliography of ...... 133 APPENDIX ...... CHAPTER -5 OPTIMAL PHYSICAL CAPITAL AND FINANCIAL STRUCTURE OF THE
OF THE SAME DECISION? 134 FIRM: TWO FACES ...... 1. Introduction_ ...... 135
2. A model of financing and investment with asymmetric information
bankruptcy and costs ...... 139 2.1 implementation The empirical ...... 146 2.2 dataset the the data The and use of ...... 151 3. The empirical results ...... 156 4. Conclusions ...... 166 Bibliography Chapter 5 of ...... 168 1 APPENDIX ...... 172 2 APPENDIX ...... 175
CHAPTER 6- CONTROLLING GROUPS, MARKET POWER, AND THE COST OF
CAPITAL IN A NON-SECURITIZED FINANCIAL SYSTEM: AN ALTERNATIVE
INTERPRETATION OF THE FIRMS' INVESTMENT DECISION 181 ......
1. Introduction- ...... 182 2. The assumptions the main of analysis ...... 183 3. A digression on the cost of capital and the "financial side of the firm" ...... 193 decision the firm 4. The problem of ...... 202 financial decisions the firm 5. The of ...... 204 "real" decisions the firm 6. The of ...... 212
7. Conclusions ...... 217 Chapter 6 Bibliography of ...... 219
7 CONCLUDING CHAPTER REMARKS ...... 223 Chapter 7 Bibliography of ...... 229 1
CHAPTER 1
INTRODUCTION
While the acknowledgments are reported separately for each single chapter of this thesis (since each chapter has involved different techniques and discussions with different people at the various stages of my work), I would like to thank here my supervisors Keith Cowling and Norman Ireland for their precious suggestions on the structure and the general organization of this work, briefly summarized here, and for their comments on a few methodological issues. Obviously none of them is responsible for any mistakes that might be found and for the views expressed here. 2
INTRODUCTION
1. The purpose of the analysis
Mainstream macroeconomic models do not usually take into account the macroeconomic implications of changes in the industrial structure of the financial or banking sectors. In addition, traditional macroeconomics is not, in general, particularly concerned with the analysis of institutions. This study attempts to investigate these two issues by means of a few theoretical and empirical analyses which are concerned with the interaction between industrial firms and financial intermediaries.
The analysis will have to deal, by definition, with aspects that would traditionally be included both in macroeconomics and in microeconomics. For this reason, and because particular attention will be dedicated to a controversial issue like the macroeconomic implications of different institutional environments, a brief discussion on the methodological premises of this study is needed, although a methodological analysis is well beyond the scope of this work. Such a brief discussion will be contained in the next-section, while section 3 will briefly describe a few contributions on which the present study is based, and section 4 describes how the work will be organized in the various chapters.
2. A few methodological premises
Dealing with the interactions between banks and industrial firms, or with the macroeconomic effects of modifications in the market structure, requires a precise choice on the well-known 3
problems of relations between micro and macro theory, of microfoundation and use (or not use) of the representative agent as an interpretative tool of individual rational behaviour.
Until the mid-1980s, the neoclassical macroeconomic literature regarded Modigliani and Miller's neutrality theorem as an acceptable simplifying hypothesis for macroeconomic modeling. Only heterodox and post-Keynesian macroeconomic literature regarded the financial structure of the firms as non-neutral, and argued that business
fluctuations could originate from financial market perturbations. On the other hand, in finance, even mainstream contributions had regarded the optimal financial structure of the firm as a central issue, long beforel such an assumption was incorporated in widely accepted micro-founded macroeconomic models (like, for instance, in
Greenwald and Stiglitz [1988], Bernanke and Blinder [1988], Bernanke and Gertler [1989]). The different aims of economic and financial
analysis did not justify per se such a relevant difference in the valuation of the firms' financial structure. Obviously the discrepancy was to be found in a different prior valuation of the
relevance of market imperfections, but the suspicion that an a priori factor might have played a role in this regard, seems to be
legitimate, as Gertler [1988] argued:
"The methodological revolution in macroeconomics in the 1970s also helped shift away the attention from financial factors, in a less direct but probably more substantial way. The resulting emphasis on individual optimization posed an obstacle".
(Gertler [1988], p. 565)
Similar reservations to the mainstream methodology had been
raised, even more explicitly, by Stiglitz [1991]:
1 See for example Jensen and Meckling [1976], Leland and Pyle [1977], Ross [1977], Myers [1984], Myers and Majluf [1984]. P 4
"Economics is, or is supposed to be, an empirical science: how could economists' views be so divergent? Were these so-called scientists studying the same economy? Were they - or should I say, looking for are we - simply ideologues justifications for our political biases, or, no less worse, technicians, taking the assumptions provided to us by our ideologue brethren, and exploring their consequences, trusting that the models we are analyzing bear some semblance to the world, because we have been told so by others! "
(Stiglitz [1991], p. 5)
"Economists have had two responses to such inexplicable phenomena. One is to suggest that because we cannot explain them, they do not exist. It is as if a biologist, finding it difficult to explain how blood can be pumped to the head of a giraffe, were to assume that it therefore must have a short neck".
(Stiglitz, [1991] p. 21)
Stiglitz' strong criticism of the mainstream methodological approach is motivated by an objection on the use of "first principles", illustrated, by the way, with an example taken from
financial economics:
"Not every piece of research has to begin at the beginning. We know that there are good reasons, based on problems of adverse selection and moral hazard, that equity markets may not function well. We also have ample empirical evidence that firms make limited use of equity markets, and event studies confirm that when they do raise additional capital through the issue of equities, stock prices are lowered significantly. It thus seems perfectly appropriate for macroeconomic studies to begin with the hypothesis that equity markets do not function efficiently. For some purposes, it may not matter what the precise source of this market failure is. "
(Stiglitz [1991], p. 10)
In connection with this point and in support of it, Stiglitz mentions the results of Debreu [1974], Mantel [1974] and
Sonnenschein [1972], [1973], showing that any set of market excess
demand functions satisfying Wairas' Law can be derived from utility maximizing individuals, which means, in other words, that the 5
rationality hypothesis does not put any relevant restriction on the observed behaviour2.
Furthermore, Stiglitz argues that the use of representative agents' models seem to contain an intrinsic paradox, since, "when all individuals are identical there is no need for trade, and hence there are no consequences of the absence of markets" (Stiglitz
[1991], p. 11). In addition, representative agents' models are not suitable for describing problems arising from information asymmetries and coordination failures, unless "a particular kind of schizophrenia on the part of the representative agent" (ibid) is assumed. However, in spite of the wide acceptance of micro foundations and representative agents as rigourous bases for macromodels, a series of assumptions commonly accepted in the microfounded models are pointed out to be "ad hoc". One of them is cash in advance, which is "obviously not binding for most transactions" (Stiglitz [1991], p. 19-20), while, for what concerns a well-known and common requirement for equilibrium to be reached, the objection is even more extreme: "What faith do we have that any propositions derived in the artificial economy in which individuals meet at most only once and there are no intervening financial
institutions have any validity for our economy? " (ibid).
2 We might briefly anticipate here that these points made by Stiglitz are relevant for the analysis of Chapter 6, on the basis of the empirical observations contained in Brioschi, Buzzacchi, Colombo [1990], Mayer [1989], [1992], [1993], showing that in the Italian financial markets hostile takeovers are extremely rare, not to say virtually absent, while the transactions concerning the control and the majority shares of a company are usually performed through private negotiations among the management of the parties interested in the transaction. For this reason it is assumed that the market for firms' control is not associated to the market for shares. The latter is regarded, in the models of Chapter 6, as a market where the firm raises external finance. 6
A more interesting point for its possible connections with other (although not explicitly mentioned) methodological approaches is made about "inexplicable phenomena", and mainly the "widespread phenomena of individuals holding dominated assets". There might not be, on the other hand, full agreement on Stiglitz' conclusion that
"it is hard for any economic theory to explain why [... ] cash management accounts, [a financial asset that should "dominate" all of the other liquid assets] [... ] did not exist twenty years ago, and it is hard for any economic theory to explain why they are not even more widespread today" (ibid). In this regard, a very relevant interpretative contribution may be provided by Williamson's [1985] contractual framework, where the assumption of rationality is not eliminated, but substituted with the well-known assumption of
"intended the "bounded rationality" or rationality" , and concept of as asset specificity is regarded the endogenous contractual outcome determined by all of the informationally relevant elements (such as timing, frequency of transaction, and many other details) affecting the decision process of the individuals3.
of Confronted with the realities of bounded rationality, the costs of planning, adapting, and monitoring transactions need expressly to be considered. [... ] Transactions that are subject to ex post opportunism will benefit if appropriate safeguards can be devised ex ante".
(Williamson [1985], pp. 46-48, also quoted in Chapter 3)
Williamson's [1985] approach, based on the relevance of transaction costs, suggests an interpretation of the behaviour of economic agents in terms of contractual relations: the relevance of
3 on this point, see also Chapter 3, which contains a more detailed description and analysis of all the points just mentioned. 7
transaction costs is also implied by those interpretations of the
Arrow-Debreu model that define commodities not only by physical,
spatial and time characteristics, but also by those elements of environmental uncertainty referred to as the "state of the world".
In Williamson's view, economies on transaction costs can be
implemented by assigning transactions to governance structures chosen among different institutional alternatives: the "classical market contracting" at one extreme, a centralized hierarchical
the firms organization at other, and mixed models -of and market organization in between. In this context, bounded rationality contrasts with the traditional approach, which suppresses the role of institutions in favour of the interpretations of firms as
"production functions", or "black boxes".
In this sense, the "Cash management accounts" mentioned by
Stiglitz might have appeared only recently and still not have a wider diffusion, to the extent that they correspond to a specific contractual outcome, responding to precise safeguard needs, determined by means of a "bounded rationality" decision process where information, planning and calculations are not costless and
timeless. To the extent that a "time dimension", or even an
"historical dimension" of the economic processes and phenomena
becomes possible, such an interpretation might explain the common-
sense observation of historically-determined or institutionally-
determined behaviour of individuals and economic systems.
These last points are relevant for the analysis of Chapter 3, which tries to interpret the phenomenon of securitization on the
basis of Williamson's [1985] approach, and provides a comparative
empirical analysis meant to show that some institutional factors 8
(namely securitization), have relevant macroeconomic implications that should be taken into account in the standard macroeconomic studies.
3. Starting points and general assumptions
After mentioning all the (well-known) criticisms of mainstream methodology, it could be objected that "destroying is easier than constructing", or, in other words, that the formulation of even the best motivated objection certainly requires less effort than the construction of a "positive", consistent and complex methodological approach. Hence the need for a brief discussion of the methodological criteria that will be illustrated and discussed.
This work will follow, in general, the methodological approach described in Stiglitz [1991], and will interpret individual behaviour and the concept of individual rationality following - at
least in spirit - Williamson [1985].
In general, this work will take Stiglitz' criticism on the use of first principles: "Hopefully our discipline is a cumulative
science. Not every piece of research has to begin at the beginning"
(Stiglitz [1991], p. 10). Following this approach means, for our purpose, that some of the models presented in this work will contain
assumptions that derive from previous results of the related
literature. The relevant literature and the specific contributions
that prove such results will be referred to, and briefly described,
but the results themselves will be regarded as starting points, or
initial assumptions, and, in this sense, will not need further
demonstration, provided that the model founded on them does not 9
explicitly violate the hypotheses of the models that proved such results. In other words, once a result has been acquired in the relevant literature, one should be allowed to investigate the implications of the introduction of such results in the hypotheses of related models.
Furthermore, when two or more contrasting assumptions are possible on the basis of different pieces of literature, an attempt will be made to qualitatively compare how the different kind of assumptions (and, possibly, the different kinds of functional links associated to them) affect the behaviour of the model being built and its results4. Obviously the validity of the conclusions of each model is limited by the prior assumptions contained in each
formalization. However, to the extent that some kinds of standard models are widely accepted in the literature (whether or not their prior assumptions or their first principles correspond to an
accurate description of reality), some informational contribution to
the debate might be obtained by analyzing the effects of introducing
some non standard assumptions in one of such "popular models". The
analysis of Chapter 2 is conducted in this spirit: the macroeconomic
effects of oligopsony in the market for credit are analyzed by
introducing a simplified (but commonly employed) formalization of
market concentration on (the demand side of) the market for credit,
in a very standard banking model, where the behaviour of the
industrial firms is microfounded. This is done by keeping well in
mind the above-mentioned criticism. In fact such a strategy of
analysis, is needed for its counterfactual value: while the effects
of market power constitute one of the main concerns of industrial
4 In particular, such a procedure will be relevant for the analysis contained in Chapter 6. 10
economics and of some heterodox macroeconomic contributions, they tend to be considered irrelevant (or, in any case, are neglected) in mainstream macroeconomic analysis. The result of such a situation is the fact that macroeconomic irrelevance of modifications in, the market structure and in the market concentration are regarded as a common modeling rule, and not as implicit assumptions, while the analysis of the macroeconomic effects of changes in the market structure is relegated to heterodox approaches. Therefore, a relevant informational contribution could be given by proving that even in a standard banking model, including the (microfounded) behaviour of industrial firms, changes in the market structure (in the case of Chapter 2, on the demand side of the market for credit) carry relevant macroeconomic implications.
In connection with these last points, a particular relevance is assumed by Stiglitz' [1991] point on dynamics and adjustment speeds.
"Any short-run macroeconomic model can be viewed as 'cutting into a dynamic process', of saying that some variables adjust more rapidly than others. More particularly, it is assumed that the present value of certain variables adjusts fully to their 'equilibrium' values. [... ] Other variables are assumed to adjust, but too slowly to worry about for short-run analysis. [... ] It is, of course, not obvious that having two categories is an adequate simplification; one might want at least to consider three categories, in which case one would discuss 'short-short run equilibrium', 'short-run equilibrium' and 'long run equilibrium'. [... ] What is clear, however, is that much of the macroeconomic theory of the past fifty years has made a set of particularly unpersuasive implicit assumptions concerning dynamics. "
(Stiglitz [1991], pp. 31-32)
Choosing what variables have to be assumed to adjust rapidly or slowly entails some prior assumptions. If, on the one hand,, such prior assumptions can be well motivated, several examples can be made of cases where such prior choices do not depend on a precise 11
observation or hypothesis on the timing of the economic process under analysis, but rather on the kind of literature, or on the kind of discipline one is dealing with. To give an example pertinent to this study (and that will be taken into account in the following chapters), there is no particular reason to assume that the financial structure is given and exogenous for the investment decision of the firm (as is usually assumed by the mainstream literature on investment), or that the level of investment is given for the, choice of the firm's optimal financial structure. Indeed, only very few contributions (and only recently) have regarded the problems of firms' investment and financial structure as simultaneous. In the case of investment and financial structure, the choice of regarding some variables as fixed and other variables as adjusting, is not even a matter of speed, but a matter of prior assumptions. The same could be said by comparing the industrial economics literature on the "deep pocket argument" (Telser [1966],
Poitervin [1989a]), and on the "limited liability effect" (Brander and Lewis [1986], Poitervin [1989b]) - which put into relation interactions between the financial structure decision with the market strategic interactions and suggests a precise causal link between financial decision and market structure - with the financial
conomics literature concerned with the minimization of the cost of capital which ignores any form of interaction with the market strategic interactions.
Relating some industrial economic results (which emphasize the relation between the financial decisions and the market power of the firms or its particular strategy) with the literature on firm's 12
investment and financial structure will be a specific concern of this work, in chapters 5 and 6.
However, the matter becomes even more complicated if one takes
Stiglitz' [1991] point on the fact that "the economy is always in the short run" and "never settles down to the mythical steady state", due to the possible presence of "sets of stochastic terms" that could constantly drive the economic system away from its
(hypothetical) equilibrium. Also, if one accepts that the behavioural functions might be subject to discontinuous modifications in their characteristics, whether such modifications are better formalized by a set of stochastic terms or by assuming some "exogenous" or discontinuous changes in the analytical forms of the functions themselves, it might often be a matter of taste and personal preference. In this sense, the modifications of standard models by introducing some (non microfounded) specific assumptions based on the results of other studies should, in principle, be acceptable.
Given all these open problems, we take Stiglitz' point in arguing that the conformity of a theory to the basic qualitative facts of the economy should be regarded as a suitable method to evaluate a theory. This means also that if a theory "fails to meet the test, there is little to be gained from the sophisticated testing of one or two of its implications, for in the end [... ] a theory must be judged by the consistency of all of its implications with the facts" (Stiglitz [1991], pp. 71-72). In Stiglitz' view, the expression "all of the implications" (of a theory) means micro- predictions as well as macro-predictions, and if even only some of the predictions are falsified (in Popper's sense), then the whole 13
theory should be "rejected or, at least, patched up" (ibid). Judging a theory on the basis of its ability to qualitatively explain some critical facts also means that econometric evidence based on the
"goodness of fit" should not be regarded as a decisive test of a model. This point also takes into account the fact that a good empirical specification might be found to be consistent with two or more observationally equivalent theoretical explanations. Obviously this does not mean that empirical analyses are not important, but simply suggests that some healthy awareness of the intrinsic logical limitations of these kind of results (which, by the way, will constitute a relevant part of the present study) be maintained.
An example of a theoretical approach that does not make use of
"the representative agent" and, instead of starting from a priori
"first principles" starts from empirical observations, is given by
Bernanke and Blinder's version of the "credit view". Such an approach starts form Bernanke's [1983] empirical analysis showing that the financial distress in 1929 amplified the effects of the great depression, whose intensity and persistence could not have been explained only by money market forces. In Bernanke's [1983] paper it is shown that in the period 1930-33, almost half of the existing American banks failed, and the remaining suffered very relevant losses, while the stock market crisis of "Black Friday"
(which was the initial event" of the crisis) determined an enormous increase in the debt burden of the firms. According to Bernanke, the banks' distress had an effect on real activity by suppressing the financial flows for some categories of firms that did not have direct access to spot financial markets and had to rely on financial intermediaries. In addition, the drastic increase in the debts of 14
-the industrial companies reduced their ability to obtain finance from the intermediaries, who would base their evaluations on the riskiness of the customers on indicators of the financial structure.
According to Bernanke, and differently from what was argued by
Friedman and Schwartz, the main role in the mechanism of propagation of the 1929 crisis would not have been played by perturbations in the banks' liabilities (i. e. in money), but rather in the banks' assets, and, in particular, by the elimination (or by the drastic reduction) of the channels by which finance was injected into the real economy. As evidence in support for his thesis, Bernanke shows that the liabilities of the failed banks and the spread between the interest rate on risky and riskless securities would significantly increase the explanatory power of the equations determining the level of output. Bernanke [1983] assumes obviously that the
"perceived" riskiness of the firms borrowing from the banking system depends on the firms' financial structure. Such an assumption, which is often formalized by introducing a risk premium (defined as a function of the leverage ratio) in the interest rate on borrowing is contained in the models of investment decision of the firm in 5. chapters 5 and 6 As mentioned earlier, Bernanke's [1983] is also one of the starting points of the well-known Bernanke and Blinder's
[1988] modification of the IS-LM framework. This model differs from the traditional macroeconomic framework by, among other things, explicitly introducing the balance constraints of the banks.
5 Such an assumption, as is well known, can actually be microfounded on the basis of an "incentive argument": loosely speaking, the higher the leverage ratio, the lower the cost of financial distress for the shareholders, and the smaller the incentive to avoid financial distress. 15
Bb +E+ LS = D(1-z) (1 where Bb is the quantity of public bonds held by the banks, E the
free reserves, Ls the credit supplied by the firms, D the deposits,
t the reserve requirements. The deposits are determined by the
liquid reserves and by the money multiplier according to the
following relation:
-++ D(i, y)=m(i)R (2
Where "D" are the deposits, "y" the aggregate income, "ill the
interest rate on public bonds, m(i) the money multiplier, "R" the
liquid reserves. Bernanke and Blinder further assume that the
portion of banks' assets invested as credit to the industrial firms
depends positively on the interest rate on loans and negatively on
the interest rate on public bonds, and is determined by the
following relation
LS = 6(r, i)D(1-z) (3
Hence Bernanke and Blinder solve the equilibrium condition on
the market for loans with respect to the interest rate on loans r, "R", so that "r" is expressed as a function of the liquid reserves
of the income "y", and on the interest rate on public bonds "i". The
resulting equation is
r= 4(i, y, R) (4
which, substituted into the IS curve, yields:
y= Y(i, 4(1, y, R)) (5 16
Equation 5 is a modification of the IS curve that contains the disturbances and the macroeconomic fluctuations determined by the attitude of the banks in their lending decisions. For example, a variation in the degree of riskiness attributed by the banks to the firms' investments, determines a shift in the "modified IS", which is defined by Bernanke and Blinder as the CC curve, i. e. the locus of simultaneous equilibrium points on the goods and credit market.
Such a macroeconomic model seems to be more consistent with the sort of "micro foundation" based on Williamson's [1985] contractual framework mentioned in the previous section because it allows for a larger degree of "asset specificity" by not aggregating all of the financial assets in a unique market (like the traditional IS-LM model does), which would cause a major loss by failing to capture the highly intrinsic contractual difference between the assets resulting from a monitoring activity (which contain safeguards against "opportunism", in Williamson's terminology) and the assets resulting from a spot market.
Being consistent with most of the theoretical foundations of the present analysis, Bernanke and Blinder [1988] play an important role in it. First of all, a supply of credit function analogous to the one of equation 3 will be used in the comparative empirical analysis of Chapter 3, meant to show the macroeconomic relevance of some institutional features, namely securitization. Secondly, a test for the reliability of the theoretical model by Bernanke and Blinder is made in Chapter 4, by empirically studying the behaviour of the free liquid reserves of commercial banks. In fact, the free liquid reserves (or the free liquidity ratio, as has been done in Chapter
interpreted "non investment" 4) can be as a decision and, in this 17
sense, is expected to react according to the willingness of banks to invest: in other words, the free liquid reserves are expected to increase when there is an increase in the degree of risk of the whole economy perceived by the banks that would reduce the willingness of banks to invest. An empirical analysis showing that the free liquid reserves (or the free liquidity ratio, which is the ratio of the free liquid reserves to the total banks' deposits) are positively correlated with some measure of "perceived risk" of the whole economy, would be consistent with Bernanke and Blinder's
[1988] model, and, more generally, with the "credit view".
4. Structure of the work,
The analysis of the second chapter is concerned with the macroeconomic effects of market concentration in the market for credit. The focus in this study is on oligopsony in the banking sector, an issue virtually ignored both by the microeconomic and the macroeconomic contributions on market power in the financial sector.
The methodological approach of the second chapter constitutes an exception in this study, since it is based on a more standard banking model with optimizing industrial firms. The reasons for this exception in the methodological approach followed in this work have been given at the beginning of section 3, and will be recalled at the end of the present section.
Most of the existing literature on market power and strategic interactions in the (supply side of the) market for credit, deals with problems of signalling in a game theoretical framework, as in the literature concerned with the "deep pocket argument" (Telser 18
[1966], Benoit [1984], Poitervin [1989a]) and the literature concerned with the "limited liability effect" (Brander and Lewis
[1986], Poitervin [1989b]). In the former it is assumed that
"strong" firms can afford a long-lasting price war because they can rely on large financial resources. In the latter, a high level of debt is regarded as a pre-commitment for an aggressive policy on the goods market. The signalling game of entry deterrence yields as a result the optimal financial structure for the incumbent and for the entrant.
Even the rare contributions concerned with the macroeconomic effects of variations in the degree of concentration in the banking sector (like Vanhoose [1985]) look at the supply side of the credit market, i. e. at the effects of market power in the banking sector.
In particular, Vanhoose [1985] looks at the central bank's ability to control monetary aggregates, under different monetary policy regimes and regulations (i. e. control of interest rate vs control of money supply and lagged reserves accounting vs contemporaneous reserves accounting).
The purpose of Chapter 2 is then to show that variations in the degree of oligopsony in the market for credit may affect the equilibrium level of investment and the monetary policy multiplier in a partial equilibrium model with banks and industrial firms. In particular, it will be shown that an increase in the degree of oligopsonistic power of the industrial firms reduces the effectiveness of monetary policy in a simplified model where the
supply of credit to the industry is summarized-by a Cobb-Douglas
function. When the banking sector is described by a more detailed
and elaborated model with three assets, behaving in accordance with 19
the portfolio allocation theory, the results are weakened, and an exogenous increase in the oligopsonistic power of industrial firms will increase or reduce the monetary power multiplier depending on the analytical form of the various functions chosen, and on the different sensitivity of the agents to the interest rates of the different assets. The reason why an orthodox banking model with optimizing industrial firms has been chosen is very simple: it is meant to stress the fact that the implicit assumption of macroeconomic irrelevance of changes in the market structure on the
(demand side of the) credit market - contained in the standard macroeconomic approach - might be questioned even by using a standard banking model with optimizing industrial firms. In other words, the purpose of the second chapter is to provide a framework that conceptually "isolates" the market power of industrial firms in the credit market and shows its macroeconomic relevance even by
following the methodological approach of the macroeconomic models that normally ignore such issues.
Chapter 3 looks at the macroeconomic effects of securitization and suggests, at the same time, an interpretation of such an
institutional phenomenon in the light of Williamson [1985]. The
analysis is again focused on the "demand side" of the market for credit and the key point is an institutional aspect connected with
the size and reputation of big firms operating in securitized
financial systems: their ability to substitute intermediated credit with recourse to the spot credit market. It is argued that
substitutability between intermediated credit and securities makes
the demand for bank credit to industry more unstable than the
relative supply function. This makes it possible to identify and 20
estimate, in a monoequational framework, a supply function of bank credit to industry. Since instability of the demand for bank credit to industry is a consequence of securitization, the identification of a supply function is only possible in a securitized financial sector, where substitutability between intermediated credit and securities is an empirically relevant phenomenon. In this sense, the empirics contained in the third chapter suggest that securitization is macroeconomically relevant. Furthermore, the stability of a credit supply function might carry some economic policy implications, to the extent that stability of a macroeconomic function could be regarded as a "relevant" and "informative" property. The analytical form of the bank credit supply function is based on the theoretical part of the paper by Bernanke and Blinder
[1988], which puts a strong emphasis on the macroeconomic effects of the attitude of banks and their willingness to lend money to the firms, affected by the degree of risk of the whole economy, such as perceived by the banking system. This specific aspect of the behaviour of banks is analyzed in Chapter 4, which contains an empirical analysis of the free liquidity ratio for commercial banks.
The free liquid reserves of commercial banks are regarded as a liquid asset associated to the non-investing decision of the bank.
Such a non-investing decision might be determined by an increase in the degree of risk of the whole economy perceived by the banks
(which is the core of Bernanke and Blinder's [1988] model), and is interpreted on the basis of the recent literature on investment decisions under conditions of investments' irreversibility and uncertainty (for instance, Dixit [1992a], [1992b], Pindyck [1991]).
This approach assumes the presence of sunk costs (in the specific 21
case of the banks they might be due to "lemon problems" and to the monitoring costs), "on-going" uncertainty concerning the profitability of future investments (which can only be inferred by the agents on the basis of probability calculus and expectations, and, in this sense introduce some elements associable to the assumption of "bounded rationality"), "relevance of decision timing"
(in other words, the investment can be delayed, allowing the bank to collect all the information affecting investment profitability, before committing its resources). The "relevance of decision timing", by putting emphasis on the material determination of the
"intendedly best" procedure, introduces again an element associable to the assumption of "bounded rationality", because it describes the intended optimization as the choice of a behavioural procedure limited by (and relative to) the effective time of choice.
Furthermore, the "relevance of decision timing" is assimilable to the expression "calculus procedure is costly" which is the basis of the assumption of bounded rationality. However, the conditions just
(i. has mentioned imply that the investor e., in this case, the bank) to take into account the presence of a positive "value of waiting for new information" before investing. In other words, the traditional "net present value" rule could be transformed into a rule suggesting that an investment should be undertaken if the net its present value of cash flow exceeds the purchase and installation least cost by an amount at equal to the value of keeping the option to invest the same resources elsewhere. In the empirical analysis of
Chapter 4, the "value of waiting" will be "captured" in the
for degree estimates by a proxy the of risk of the whole economy by banks, introduced such as perceived the in a standard model of 22
the free liquidity ratio for commercial banks. In this sense, the empirical analysis of Chapter 4 is meant to provide some evidence in favour of the connection between "risk perceived by the banks" and
"banks' willingness to invest", which is highly relevant and, at the same time, exogenously assumed in Bernanke and Blinder [1988].
Having said that, the key feature of Bernanke and Blinder
[1988] - as well as most of the contributions of the macroeconomic
"credit view" - is the willingness of banks to lend (in general affected by the "perceived" degree of risk of the whole economy), what about the "microeconomic" level? How does risk affect the cost of capital and the investment decision? A fruitful way to approach such a problem is, in our opinion, the formalization of a model where the decisions concerning the financial structure and the investment of the firm are simultaneous. This is precisely the scope of Chapter 5, which contains an empirical analysis of the firms' investment decision founded on a theoretical model where the relevance of the financial structure - based on the assumption of asymmetric information - is summarized by the presence of a "risk premium" in the cost of borrowing, and other standard assumptions taken from models of finance with market imperfections. Although the formalization of the investment and financial structure choice as a simultaneous decision for the firm determines a significant informative contribution and puts the interactions between industrial firms and the financial sector in a more general light,
degree it increases the of complexity of the models under consideration. As a consequence, some simplifying assumptions are
for often necessary, the sake of tractability of the models, at the loss cost of some of information. Such losses of information are the 23
main concern of Chapter 6, which contains an analysis of the implications of a few alternative hypotheses on the link existing between the cost of capital, the market structure and the profit margins. All of the alternative hypotheses taken into consideration are motivated by precise results of the industrial economics literature concerned with the connection between firms' strategic interaction and financial factors like the choice of the firm's financial structure and the determination of the cost of capital. In many parts the analysis is deliberately qualitative, in order to put into evidence the qualitative changes in the results determined by the different hypotheses.
Finally, Chapter 7 contains a few conclusive remarks and an attempt to assess the informative contributions given by the analyses contained in the different chapters. 24
Bibliography of Chapter 1
Benoit, J. -P., [1984], "Financially Constrained Game in a Game with Incomplete Information", Rand Journal of Economics, 15, pp. 490- 499.
Bernanke, B. S., [1983], "Non-Monetary Effects of the Financial Crisis in the Propagation of the Real Depression", American Economic Review, vol. 73, pp. 257-276.
Bernanke, B. S., Blinder, A. S. [1988], "Is It Money or Credit, or Both, or Neither? - Credit Money and Aggregate Demand", American Economic Review, vol. 78, pp. 435-451.
Bernanke, B. S., Gertler, M. [1989] "Agency Costs, Net Worth and Business Fluctuations", American Economic Review, vol. 79, pp. 14-31.
Brander, J. A., Lewis, T. R., "[1986], Oligopoly and the Financial Structure - The Limited Liability Effect", American Economic Review, vol. 75, pp. 956-970.
Brioschi, F., Buzzacchi, L., Colombo, M. G., "Gruppi di Imprese e Mercato Finanziario", Roma, La Nuova Italia Scientifica.
Debreu, G. [1974], "Excess Demand Functions" Journal of Mathematical Economics, vol. I, 1974, pp. 15-21.
Dixit, A., [1992a], "Irreversible Investment with Uncertainty and Scale Economy", mimeo, April, Princeton University.
Dixit, A., [1992b], "Investment and Hysteresis", Journal of Economic Perspectives, Winter.
Gertler, M. [1988], "Financial Structure and Aggregate Economic Activity: An Overview", Journal of Money, Credit and Banking, vol. 20, pp. 559-588.
Greenwald, B. C., Stiglitz, J. E. [1988], "Imperfect Information, Financial Constraints, and Business Fluctuations", in "Financial Constraints, Expectations and Macroeconomics" Oxford, Oxford University Press.
Jensen, M., Meckling, W., [1976], "Theory of the Firm: Managerial Behaviour, Agency Costs and Ownership", Journal of Financial Economics, vol. 3, pp. 305-360.
Leland, H., Pyle,. D., [1977], "Information Asymmetries, Financial Structure, and Financial Intermediation", The Journal of Finance, vol. 32, pp. 371-398.
Mantel, R. [1974], "On the Characterization of Aggregate Excess Demand", Journal of Economic Theory, vol. 7, pp. 348-353. [1989], "The Mayer, C., Influence of the Financial System on the British Corporate Sector", mimeo prepared for the Conference 25
"The Separation of Industry and Finance and the Specialization of Financial intermediaries" at the Universitä Bocconi, Milano, Italy.
Mayer, C. [1992], "Corporate Finance", mimeo, forthcoming in The New Palgrave Dictionary of Economics and Finance
Mayer, C., [1993], "Ownership", mimeo, University of Warwick, U. K.
Myers, S. C., [1984], "The Capital Structure Puzzle", The Journal of Finance, vol. 39, pp. 575-592.
Myers, S. C., Majiuf, N. S., [1984], "Corporate Financing and Investment Decisions When Firms Have Decisions that Investors Do not Have", Journal of Financial Economics, vol. 13, pp. 187- 221.
Pindyck, R. [1991], "Irreversibility, Uncertainty and Investments", Journal of Economic Literature, September, 29.
Poitervin, M., [1989a], "Financial Signalling and the 'Deep Pocket' Argument", Rand Journal of Economics, 20, pp. 26-40.
Poitervin, M. [1989b], "Collusion and Banking Structure of a Duopoly", Canadian Journal of Economics, XXII, no. 2, pp. 263- 277.
Poitervin, M., [1990], "Strategic Financial Signalling", international Journal of Industrial Organization, 8, pp. 499- 518.
Ross, S., [1977], "The Determination of the Financial Structure: the Incentive-Signalling Approach", Bell Journal of Economics, vol. 8, pp. 23-40.
Sonnenschein, H. [1972], "Market Excess Demand Functions", Econometrica, vol. 40, pp. 549-563.
Sonnenschein, H., [1974], "Do Walras' Identity and Continuity Functions Characterize the Class of Community excess Demand Functions? " Journal of Economic Theory, vol. 6, pp. 345-354.
Stiglitz, J. E., [1991], "Alternative Approaches to Macroeconomics: Methodological Issues and the New Keynesian Economics", NBER Working Paper no. 3580.
Telser, L. G., [1966], "Cutthroat Competition and the Long Purse", Journal of Law and Economics, 20,1966, pp. 26-40.
Vanhoose, D. D., "Bank Market Structure and Monetary Control", Journal of Money, Credit and Banking, vol. 17, pp. 298-311.
Williamson, 0., [1985], "The Economic Institutions of Capitalism" New York, The Free Press. 26
CHAPTER 2
MONETARY POLICY WITH OLIGOPSONY IN THE MARKET FOR CREDIT.
*I am very grateful to Giancarlo Bertocco, Carluccio Bianchi, Keith Cowling, Gianluca Femminis, Umberto Galmarini, Norman Ireland, Adrian Jimenez Gomez, Marcello Messori, Neil Rankin and David Vanhoose for helpful discussions and suggestions at different stages of this work. All mistakes are obviously mine. 27
MONETARY POLICY WITH OLIGOPSONY IN THE MARKET FOR CREDIT.
1. Introduction.
The purpose of this paper is to analyze the effects of the market power of industrial firms in their relationship with banks.
The role of banks is still a controversial theoretical issue: it has been explained as a response to the financial markets' imperfections and incompleteness (Diamond [1984]), which means that financial intermediaries would be unnecessary and irrelevant in a complete market system ä la Arrow-Debreu.
Although information asymmetry is certainly a key feature in understanding the mechanisms of the market for credit, one cannot exclude that market power in itself also affects the cost and supply for bank credit.
The effects of the banks' market structure on monetary policy have been analyzed in an important contribution by VanHoose [1985], with respect to the effects of market power on the central bank's ability to control monetary aggregates, under different monetary policy regimes and regulation (i. e. control of interest rate vs control of money supply, and lagged reserves accounting vs contemporaneous reserves accounting).
This paper will analyze the effects of oligopsony on the credit market, and on the transmission mechanism of monetary policy. This will be done, first in a simplified framework, where the credit supply to the (oligopsonistic) industrial firms is described by a reduced form of the monetary sector. Secondly, a generalization of the analysis will be done by describing in greater detail the behaviour of the banking sector. For this purpose the effects of an 28
exogenous modification in the market structure will be introduced in a model similar to the one by Hörngren [1985], which is basically a model of portfolio allocation ä la Tobin and Brainard [1963].
Section 2 contains a simplified model of partial equilibrium where the industrial firms are oligopsonistic in the credit market.
In this over-simplified model, the monetary sector is described by a constant elasticity credit supply function whose arguments are the
(unique) interest rate, and a generic parameter describing monetary policy. Within this simplified framework an increase in the degree of concentration in the industrial sector (i. e. in our case an increase in the oligopsonistic power of industrial firms in the credit market) will reduce the effectiveness of monetary policy.
Section 3 contains a model of the banking sector, consistent with the portfolio choice theory. In this model, an increase in the degree of competition (a reduction in the degree of concentration) of the industrial sector on credit market, has an expansive effect, increasing the optimal level of the investments of the industrial firms. The same model is also employed to attempt an analysis of the results of the simplified case of section 2. Those results are no longer general because, in a more general framework, variations in the market power of industrial firms have an indirect effect on the
"sensitivity" to the interest rate of the "optimal", level of investments of the industrial firms that might have an opposite sign to the one of the direct effects.
Since the attention is here focused on a very specific point
(namely the oligopsonistic power of industrial firms on the market for bank credit), some simplifying assumptions will be made in order to "isolate" the interaction between banks and industrial firms from 29
other forms of interactions involving the goods market. In particular, while assuming oligopsony on the market for bank credit, the goods market will be assumed to be perfectly competitive.
Such an assumption, apart from simplifying dramatically the structure of the model, might describe those institutional contexts
(relatively frequent in continental Europe) where large industrial firms are exposed to international competition in the goods market while enjoying some market power in the internal market for credit.
Such a situation might be determined by the regulatory limitations in the banking sectors, concerning investments in foreign assets.
2. Monetary policy in a simplified model with oligopsonistic
industrial firms in the credit market.
In this model the labour market will not be considered. The analysis will be focused on the capital market.
Let us assume that in the market there are N equal enterprises producing a unique homogeneous good and using a production process whose characteristics may be described by a standard Cobb-Douglas production function
y= g(k) = Bka; with 0 N, the number of enterprises, is fixed, while "n", the number of firms, can vary. In this way, letting "n" vary, the scale of the economy will not be affected. Only the degree of concentration, or 30 in other words the number of enterprises per firm N/n, will be affected. Let us also assume that the life of physical capital be one period only, so that "k" is both the capital stock of each enterprise and the investment of each enterprise, and that the firms finance their investment only with borrowed money. We have then K= Nk where the aggregate stock of capital also corresponds to the aggregate level of investments. Let us assume that the reduced form of the monetary sector of the economy may be described by the following constant elasticity credit supply function: S(r, 0) = Arß6 where "A" is a constant, "r" the interest rate on loans granted to the enterprises, "ß" the interest rate elasticity of the supply function of loans, "8" a generic parameter describing the monetary policy such that the higher 8, the more expansive the monetary policy. In this simplified framework, the effect of the monetary policy on investments of industrial firms will be given by: dK/dO = d(Nk)/d9. In order to verify whether variations in the degree of concentration in the industrial sector exert any influence on the effectiveness of monetary policy, we have to see how dK/d8 varies when "n" varies. We have therefore to calculate the derivative of dK/d8 with respect to "n". Since N, the number of enterprises, is 31 fixed, and the firms are identical, we can equivalently look at the multiplier d(dk/dG) do If this multiplier is positive (given that the multiplier dk/dO of the monetary policy is positive, as we will see below), then an increase in the degree of competition (decrease in the degree of concentration) in the industrial sector in the credit market will increase the effectiveness of monetary policy. Let us now analyze the behaviour of the firms. Let us assume, on the basis of the scenario pictured in section 1, that the credit market can be described by an oligopsony ä la Cournot, while the industrial firms operate on perfectly competitive goods markets. The problem of the representative firm can be described as: N max n= . [g(k) - rk] n N s. t. k+ K' = S(r, 9) n where it is the profit, and K' the capital of all the other firms. The constraint of the optimization problem of the firm can be rewritten as: k= [S(r, 8) - K']n/N The following first order conditions can be obtained: Sn N [(69(k)/6k 8S(. ) n _" - r) - kl =0 Sr n Sr NJ 32 hence 8g (k) kN/n =r+ Sk 6S(. )/6r since in equilibrium we have S(r, 8) = Nk, then we can write 1 aBka-1 -r(1+ nß hence, solving for k: 1/(a-1) r1 k =L (1 + 1/nß)rI (1 Ba from S(r, 8) = Arß6 = Nk we get r= (Nk/AO)1/ß substituting for r in (1: 11 1/(a-1) k=r(1+-) (Nk/AO)1/ß] L Ba nß which, solving for k, determines: ß1 j' 11 ß(a-1)-1 1-ß(a-1) k=I(1+) (N/A)1/ßý e (2 L Ba nß At this point we are able to calculate the monetary policy multiplier and its derivative with respect to "n", which gives us information on the variation in the effectiveness of monetary policy 33 determined by a variation in the degree of concentration in the industrial sector. 0_ p (ac-1) 11 ß(a-1)-1 1-ß(a-1) dk/dO = [1+1/(nß)] (N/A)1/Pl 8 1-ß(a-1) Ba (3 ý(2-ac)+1 d(dk/d6) -ß 1 'j ß(a-1)-1 [1+1/(nß)] (N/A)1/ßJ " do [ß(a-1)-1] 2L Ba -O(a-1)1-ß(a-1) " (1/Ba)(N/A)1/ß [-1/(n2ß)] 0 (4 From expressions 3 and 4 it is easy to verify that dk/d6 >0 and d(dk/dO)/dn >0. The first inequality simply shows that an expansionary monetary policy (described by an increase in the value 0) increases the level of investments of the industrial firms. The second inequality shows that the multiplier of the monetary policy is an increasing function of "n". This means, in other words, that an increase in the degree of competition in the industrial sector will increase (in this simplified framework) the effectiveness of monetary policy. This is due to the fact that an exogenous increase in the degree of competition in the industrial sector (in other words a reduction in the oligopsonistic power of the firms in the credit market), by reducing the difference existing between marginal productivity of capital and interest rate, implies, ceteris paribus, an increase in the optimal level of investments of the industrial 34 firms. Such an expansionary effect also causes an increase in the absolute value of the monetary policy multiplier. We can see from equation 4 that the result does not depend on the assumption of decreasing returns to scale in the production function, but, it rather depends on the analytical form of the credit supply function, although the use of a Cobb-Douglas is common in literature, even to describe e credit supply function. However, the above result seems to be consistent with the empirical evidence supplied by a recent contribution by Gertler and Gilchrist [1991] where it is shown that monetary policy has much stronger and more significant effects on small firms than large firms. On the other hand, removing some of the simplifying assumptions of the model weakens the results, as we will see in section 3. 3. A generalization: banking sector and households. In this section we extend the framework described in section 2 by introducing a more elaborate banking sector. The model contains a more detailed description of the monetary policy, no longer summarized by a generic shift parameter 0 in the credit supply function. The results obtained in the simplified model of the previous section will be weakened because an exogenous variation in the degree of competition in the industrial sector modifies the sensitivity (with respect to the interest rate) of the industrial 35 firms' optimal level of investment. As in the previous section, we will assume that capital only lasts for one period, and corresponds to the level of investments. The theoretical background for the behaviour of the banking sector in our model is provided (apart from a few relevant modifications that we are going to introduce) by Hörngren [1985]. As earlier mentioned, Hörngren's model is basically a model of banking and portfolio allocation ä la Tobin and Brainard [1963]. The agents operating in the model are industrial firms, banks, monetary authorities and households (unlike in Hörngren [1985], the "non-bank" financial intermediaries are not consider here), and all variables are expressed in real terms. The industrial sector is constituted, as before, by symmetric enterprises, producing the same homogeneous good, with a strong-market power with respect to the banking system. All the investments of the industrial enterprises are financed by loans obtained from banks. We consider a short run static partial equilibrium model, i. e. we do not consider the feedback from the non-financial variables to the financial sector. Following Hörngren [1985] (and most of the literature on banks' behaviour), we will assume, for what concerns the financial assets demand and supply functions, that the partial derivatives with respect to their own interest rates are greater, in absolute values, than the cross derivatives. 3.1 The Monetary Authority In this model, given the level of public debt (which is assumed to be fixed and exogenous), the monetary authority implements open 36 market operations by purchasing public bonds from households and banks. Apart from open market operations, the central bank could use another instrument of monetary policy, the level of reserve requirements, as in Hörngren. In what follows we will, for simplicity, consider only open market operations, since an analysis focused on the use of reserve requirements as an instrument of monetary policy would give similar results, and open market operations are more commonly employed as an instrument of monetary policy. Let us assume that the characteristics of the economy are such that the payments are entirely performed using deposits. The money base of the economy is then given by bank reserves. Considering that we are dealing with a partial equilibrium model, where we neglect the feedback between the real sector and wealth, if we neglect the foreign sector, the balance sheet of the central bank can be simplified'as follows: BC =R (5 where BC is the amount of public bonds held by the central bank and R represents the reserves of the commercial banks. The reserves that the commercial banks hold at the central bank include reserve requirements and free reserves. The central bank, while varying the value of BC, performs open market operations, and controls the money base. Therefore the money base, BM, will be equal to BC. The balance sheet of the central bank can be regarded as a sort of budget constraint, which has to be satisfied ex post, given the behavioural relations of the model. The public debt BT (exogenous) will be given by BT = BP + Bb + BC (6) 37 or, equivalently BT = BP + Bb + BM (6' where Bb is the quantity of public bonds held by private banks. 3.2 Households This model does not include labour6, but only capital as a production factor, owned by households, which receive all the income produced. The private sector has a wealth endowment which enters the demand functions for financial assets. For simplicity, in the absence of the foreign sector, the financial wealth will be equal to the public sector debt, which is exogenous and fixed. We will limit our analysis to the impact of a variation in the degree of concentration in the industrial sector on the optimal level of investments, and on the monetary policy multiplier. The financial wealth is assumed to be constant. Let us assume that households do not lend funds directly to industrial firms, but can only choose between investing in public bonds or bank deposits. This last assumption (which also appears in Hörngren [1985]) is theoretically justifiable by the fact that banks enjoy scale economies in collecting information, or by the fact that the size of negotiated loans may be large with respect to the financial wealth of the single individual). It may be considered as an extreme description of those institutional contexts (such as that of Japan and continental Europe), where financial funds are predominantly intermediated by the banking sector. 6 We could make a "ceteris paribus" assumption, and imagine that the labour is included in the constant "B", which appears in the function production y=Bka . 38 We can further assume that the demand for bank deposits of the public is given by a "fixed" component, which can be thought of as money transactions, and taken as exogenous, and a "variable" component which can be thought of as an increasing function of the interest rate on deposits and a decreasing function of the interest rate on public bonds. If we assume that one period of time passes between the moment when the industrial firms obtain the loans and the moment when the income determined by the production process is available for the households, we can express the "transactions" component of money demand as a function of Yt_i 7. Therefore, the households' demand for bank deposits can be described as follows: DP = C(Yt-1) + DVP(rB, rD, rL, W) i7 where C(Yt-1) is the "fixed" component of the deposits, which we can assume to be exogenous and non-remunerated; DVP(rB, rD, W) is the "variable" component, function of the interest rates, DP is the total amount of deposits, W the wealth, rB the interest rates on public bonds, rD the interest rate on bank deposits, and rL on bank loans. Consistently with the portfolio allocation theory (and with Tobin and Brainard [1963] model, in the version presented by Hörngren [1985]) we can now define the other demand functions for financial assets and liabilities by the public: 7 This last assumption is not strictly necessary, since, for the purposes of the present paper, it would be equivalent to assuming that the "transactionary" component of money demand is exogenous. However, it might be more convenient to keep the money demand function in this form, (by assuming that the transactionary component of the money demand is pre-determined rather than exogenous), in order to be aware of the potential extensions of the model, once we have taken into account the feedback between the real sector and the financial sector by making explicit the link between the real income and the process of wealth accumulation. 39 +--+ BP = BP(rB, rD, rL, W) (demand for public bonds) (8 -+-+ DP = DP(rB, rD, rL, W) (demand for bank deposits) (9 +-++ Ld = Ld(rg, rL, rD, W) (demand for banks' loans by the public) (10 The budget constraint of the public is: W= BP + DP - Ld (11 as we have said, we will have W= BT 3.3 The Banking Sector The banks finance themselves by issuing deposits. Let us assume that the total amount of deposits is given by the sum of (non remunerated) current account deposits and (remunerated) saving deposits. Let us also assume that the banks are willing to issue any quantity of non remunerated deposits demanded by the public. Therefore we will have: D= Cb + DV. (12 where Cb is the non remunerated component of bank deposits, DV the remunerated one which depends on the interest rates. The remunerated component of deposits,. DV are the money market liabilities that banks issue in order to provide themselves with loanable funds, according to the assumption of "active liability management" previously mentioned. The funds collected can be invested (once the reserve requirement has been satisfied) in loans, (non remunerated) free reserves, or public bonds. 40 If, for the sake of simplicity, we neglect the shares, the balance constraint of banks is given by: LS +R+ Bb =D (13 LS = supply of credit; R= commercial banks' liquid reserves at the central bank; D= bank deposits. The banks' demand for public bonds will depend positively on the interest rate on public bonds, negatively on the interest rate on loans (which represent an opportunity cost) and the interest rate on time deposits. We have, therefore: +-- Bb(rB, rD, rL) (14 Public bonds, unlike loans, can be sold on a spot market. This implies that public bonds held by commercial banks can be regarded both as a form of portfolio investment and a partially liquid asset, to the extent that it is traded on a "spot" market instead of a "customer" market. We say "partially" because, although Bb is traded on a "spot" market, we cannot exclude, for the sake of generality, that it carries some transaction costs. On the other hand, we can assume that commercial banks hold free (non constrained) liquid reserves for transactionary purposes, and in order to be able to satisfy unexpected requests of deposit reimbursements by the public. Therefore, since Bb is, for the bank, also a liquid asset, we may think that the transaction costs associated with Bb could determine some kind of hierarchy in the allocation of liquid assets between R and Bb, and we can expect that 41 the opportunity cost of the free liquid reserves might be mainly represented by the interest rate on loans. The reserves R include the reserve requirements and the free liquid reserves. Therefore, if we define: qp = reserve requirements coefficient; q'= free liquidity ratio of the commercial banks; we can write: R= q0'D + q'"D = = (q0 + q')"D We will assume, for simplicity, that q' behaves as follows: q' = q'(q0, rL) since, assuming some form of "hierarchical" funds allocation between the two liquid assets R and Bb (determined by the fact that Bb, unlike R, brings some transaction costs), we can regard rL as the main opportunity cost for R, and we may regard as negligible the effect of the interest rate on public bonds. Furthermore, as it is argued in Hörngren [1985], the higher the reserve requirements, the more costly it is for the bank to hold free reserves. Considering equation 5, we can write: BM =R= q(qo, rL)D; (15 with q= q0 +ql and 0 then: D= BM/q("); and defining q(. ) = 1/ý(") (16 we have: 42 D(") = BM"c(") (17 and, obviously, >18 In accordance with Hörngren's [1985] "active liability management" assumption, the banks issue remunerated time deposits in order to provide themselves with loanable funds. The amount of time deposits that the banks are willing to issue is then decided simultaneously with the supply of credit to the industrial firms. This means, that for the purposes of our model, the remunerated component of deposits DV, is directly linked to the supply of loans by the bank. While Hörngren [1985] assumes that (due to the market structure of the banking sector) the relevant variable affecting the supply of banks' loans and the remunerated deposits is the (constant) spread "rL-rD", we will simply assume here that there is a generical functional link between the interest rate on bank loans rL and the interest rate on ("variable") remunerated deposits rD. In other words, we are saying that a functional link rD(rL) exists and is determined by the fact that the remunerated deposits are issued by the banks in order to provide themselves with loanable funds, given 8 in a previous version of this paper we assumed q= q(rL, rD, rB, q0) where rL, is the interest rate on bank loans, rB the interest rate on public bonds, and rD the interest rate on deposits. This seems to be a less restrictive assumption. However, as will be clear later, some additional restrictions had to be taken, in order to obtain a bank credit supply function behaving consistently with the portfolio allocation theory. In particular, we would have to assume: I(Sý/SrB)BMj < l6Bb/8rBl. Having made this assumption, the results would be equivalent to those of the present model, but the calculations would be more complex. 43 the (perfectly competitive) market structure of the banking sector, and given the presence of administrative costs for the banks (which prevents, in equilibrium, rD and rL to be equal). Our assumption is not very different from the one made by Hörngren, who, having defined LS((rL-rD), g0) as the supply of banks' loans, assumes SLS(")/rL =- öLS(")/rD" In our case, having defined rD = rD(rL) (18 we assume, for simplicity9 örL/Srp =1 (18". A similar assumption, although slightly more restrictive, is found in Modigliani and Papademos [1980] and [1987], where it is simply assumed that in a perfectly competitive banking sector, neglecting the administrative costs, the equality rD=rL holds. For what concerns the nonbank agents, their behaviour functions will still include the interest rate rD on "variable" deposits as an explanatory variable, but we will have to keep in mind the functional link rp = rD(rL). In other words, we will have: +--+ BP = BP(rB, rD(rL), rL, W) (demand for public bonds) (8' -+-+ DP = DP(rB, rD(rL), rL, W) (demand for bank deposits) (9' +-++ Ld = Ld(rB, rL, rD(rL), W) (demand for banks' loans by the households) (10' For what concerns equations 9 and 10, the sign of the partial derivatives with respect to rL is unambiguously defined, if we 9 The following assumption of equation 18" could actually be weakened, but it has been expressed in this form for the sake of simplicity. 44 consider that we have assumed that the partial derivative with respect to their own interest rate is greater, in absolute value than the cross derivatives, and we take into account equation 18". For what concerns the banks, we can rewrite equation 17 as follows D. = BM"-I)(rL) (17' Like in Hörngren [1985], we assume that the "variable" component of deposits issued by commercial banks depends on the profitability of bank loans: SD(") 6.1,(") _" BM >0 (19 SrL örL In equation 19, the interest rate rD does not appear because the bank "recognizes" the link existing between rL and rD, and equation 19 express the amount of deposits (consistent with the banks' maximization of profits and to the unit cost rD of the "variable" component of deposits, "linked" to rD) that the banking system is willing to issue, given the fact that banks issue remunerated deposits only to provide themselves with loanable funds, to be supplied on the loans market, remunerated at the rate rL. Introducing the functional relation rD(rL) into the banks' demand for public bonds does not create any particular problem: +-- Bb(rB, rD(rL), rL) (14' budget The banks' const: raint can also be written as follows: Bb + LS = [4, (") - l]-BM (13'. 13' implicitly Equality defines the supply of banks loans: -+++ +- Ls = LS(rB, rL, BM) =[ (rL) - 1]-BM - Bb(rB, rL) (20. 45 Considering equation 20 with equation 10, we can define the function S("), which represents the supply of bank credit available to the industrial firms: -++++-+- S(rB, rL, BM") =[ (rL) - l]-BM - Bb(rB, rL) - Ld(rB, rL) (21 From equation 19, we have oLd/SrL <0. This can be easily verified if we look at equation 10', and we consider two assumptions that we have made earlier: the first is that the derivatives of the demand and supply functions with respect to their own interest rates are greater, in absolute value, than the cross derivatives; the 10, second is 6rD/8rL=1 3.4 The Industrial Sector The industrial sector has the same characteristics as the one described in the simplified model of section 2. The differences lie in the behaviour of the banking sector, which is described here by a more complex model, instead of being captured by a reduced form, like in section 2. The optimization problem of the representative enterprise is the following: max n= (N/n)( f(k) - rLk) 10 In a previous version of this paper we assumed that the banking system only lent to the industrial firms. In this case, equation 21 would have been simplified as follows: -++++- S(rB, rL, BM") = ['(rL) - 1]"BM - Bb(rB, rL) the result would have been equivalent, but we have preferred to remove this restriction and assume that households can borrow from the banking sector, since this does not excessively complicate the model. 46 s. t. N "k+ K' = S(") n or, equivalently, k=( S(") - K') n/N where K' is the capital of the remaining n-i firms, S(") is the supply of credit (loans supplied by the banking system), it the profit, rL the interest rate on the loans to industrial firms. Assuming that the second order conditions are satisfied, the first order conditions are the following: Nr 8g(k) SS(. ) n Sn/SrL = - rL )- ki =0 n L( 6k örL N hence 1 kN/n 6g(k)/6k - rL -=0 n SS(. )/SrL Since, for symmetry, we have S(. ) = Nk, then we can write: 1 S(. ) aBka-1 - rL -= fl(k, rL, n) =0 (22 n SS(. )/SrL or, solving with respect to k: 11 S(") 1/(a-1) k= (rL + )] (23 Ba n SS(")/örL Expressions (22) and (23) are then F. O. C. of the optimization problem of the representative firm. They describe the relation between the marginal productivity of capital and rL (or, equivalently, a relation between k and rL), in a context 47 characterized by oligopsonistic power of industrial firms on the market for credit. 3.5 Equilibrium Conditions Since the money base is exogenous, there are three relevant financial assets in our model: bank loans, public bonds and bank deposits. Moreover, we must add condition (22) to the equilibrium conditions of the assets markets. Condition (22) defines a relation between the marginal productivity of capital and the interest rate on the loan market, and (given the elasticity of output with respect to capital in the production function) a relation between the optimal level of investments and the interest rates. It contains the information concerning the oligopsonistic power of industrial firms in the credit market. Since the market for bank credit is described by an oligopsony "ä la Cournot", we cannot properly define a demand function for capital, from the industrial firms, but we have an optimizing decision taken by the firms simultaneously to the decision of the monetary authority concerning the value of the monetary policy instrument, given the behaviour functions of the agents operating on the assets markets. Therefore the value for k chosen by the individual enterprise (and as a consequence the aggregate value K= Nk, since in an oligopsony ä la Cournot the behaviour of the firms is symmetrical) is the level of investments resulting from the F. O. C. of the optimization problem of the firm. k can be intuitively thought of as the equilibrium value in the game 48 describing the behaviour of the "n" oligopsonists, given the banks' credit supply available to industrial firms. In our model we have three markets: loans, deposits and public bonds, and three interest rates: rg, rL and rD. Equation (22) determines the relation existing between k, rL, and the other interest rates which appear in the supply of bank credit. Since the (competitive) market structure in the banking sector and the assumption of "active liability management" determines rD through equation 18, and since W, BT, and yt_i are given, we have actually three unknowns: k, rB, rL. 1 S(. ) aBka-1 - rL -=0= fl(k, rL, rB, n, BM) n SS(")/SrL (22 Considering that the aggregate investments Nk are equal to the banks' credit supply available to the industrial firms S("), we obtain the equilibrium condition for the market for bank loans: +-+ f2(k, Nk - S(rL, rB, BM) = rL, rB, BM) =0 (24 Considering equations 14', 8, and 6, we obtain the condition of equilibrium on the market for public bonds: +-+-+ Bb(rB, rL) + BP(rB, rL, W) + BM - BT = f3(rL, rB, BM) =0 (25 Combining equations 9' and 17' we obtain the equilibrium conditions for bank deposits: -+-+++ W) D(rL, BM) f4(rL, DP(rB, rD(rL), rL, - = rB, BM) =0 (26 49 Condition (22) is "microeconomic", since it is obtained from the F. O. C. of the optimization problem for the industrial firm. Equations (24), (25) and (26) are, on the contrary, "macroeconomic" conditions. Given equation 11, defining the wealth constraint of the private sector, equation 5, defining the budget constraint of the central bank, and equation 6', one can see, after doing some substitutions, that equations 25 and 26 are not independent. We will therefore drop equation 26, and we will work on consider only the remaining three. 3.6 Comparative Statics We must now first analyze the effects of a variation of the exogenous variables BM and "n" (money base and number of oligopsonistic firms in the market) on the endogenous variables of the system. Then, after calculating the value of the monetary policy multiplier dk/dBM, we can see how a variation in "n" affects this multiplier, taking into account the direct effects of a variation of "n" on the money multiplier dk/dBM, as well as the "indirect" effects, induced through a modification (induced by a variation in "n") in the equilibrium levels of the interest rates, and in the higher order derivatives of the functions relevant for the comparative statics. Assuming that the vector function F (composed of the four functions fl, f2, f3, f4) satisfies the conditions of the implicit function theorem, we may implement a comparative static analysis considering the effects of a variation in the monetary policy 50 instrument BM and in the number of firms "n", on the endogenous variables of the system. We consider equations 22,24,25, and differentiating them at the equilibrium we get the following system: 6fl 6fl 6fl 6fl 6fl dk -- 6k SrL SrB OBM ön dBM öf2 6f2 8f2 6f2 drL = -0 Sk örL 8r13 SBM do 6f3 öf3 6f3 0 drB -0 SrL SrB ÖBM (27 We assume that system 27 is stable; the appendix contains the algebraic details concerning system 27, and an explanation of the signs of all the relevant partial derivatives. Let us define A the on the left hand side of system 27, (-1 the determinant the matrix , of matrix A, and aij the element of matrix A placed in the i-th row and j-th column. We have then: +- 1 [- '5f 3' ] +- drH/dn = a21 a32 >0 (28 fý, Sn +- 1r sfl ] ++ drL/dn = {- a21 a33 }>0 (29 L-ön 1 [- Sfl I- [- Sfl ]}>0 dk/dn ={ a22 a33 a23 a32 (30 on On 51 Inequality 30 is true because we have assumed that the derivatives with respect to their own interest rates are greater, in absolute value, than the cross derivatives. As shown in the appendix, this implies that 1a221 > ja231 and 1a331 > 1a321 In the appendix it is also shown, given our assumptions, that J1. > 0. Form 28,29 and 30, that an exogenous increase of the degree of competition (a reduction of the degree of concentration) in the industrial sector increases the interest rate on loans, public bonds and the optimal level of investments for the industrial firms. This happens because a reduction in the degree of concentration in the industrial sector, reducing the spread between marginal productivity of capital and interest rate on loans, leads to an expansionary effect. Let us now consider the effect of monetary policy. 1-+++-+ {[a11(ß(")-1)a33 l+ drL/dBM = [a13 a21 (-1)] - [all 823 (-1)]+ +++ C öfl a21 a333 <0 (31 ÖBM +--++- 1 [- 8fl ] drB/dBM = ([all a22 (-1)] + a21 a32 + SBM (-1)] [all (c(")-1) -[a12 a21 - a32]} <0 (32 52 ++-+-+ 1 r- öfl 1 dk/dBM ={ a22 a33 + [a12 a23 (-1)] + L1, L SBMJ , +-+--+ + [a13 (ý(")-1) a32 ]- [a13 a22 (-1)] - [a12 ('(")-1) a33] + öfl a23 832 } (33 CSBM The sign of 33 is uncertain. When 33 is negative, the monetary policy will have perverse effects, i. e. and increase in the money base will reduce the level of investments instead of increasing it. The monetary policy will not have perverse effects when the positive addenda of 33 are greater than the absolute value of the negative addenda, i. e. when: [- Sfl ] a12 a23 (-1) - [a12 (4(")-1) a33] - a23 a32 > SBM r Sf]. I > a22 a33 + a13 (ý(")-1) a32 + a13 a22 (34 L-SBM At this point, the most obvious strategy would be to perform the analysis for two different cases: effective and perverse monetary policy. However, if the monetary policy has perverse effects (i. e. if condition 34 is not met), it seems to be quite pointless to investigate how and wether market structure affects it, because, in any case, a monetary policy with perverse effects is not advisable. Therefore, in what follows, the analysis will be implemented only for the case of monetary policy effectiveness, i. e., when condition 34 is true. This means that Sk/SBM > 0, i. e., the multiplier of monetary policy 33 is positive. 53 Now let us analyze the effects on the monetary policy multiplier of an exogenous increase in the degree of competition (decrease in the degree of concentration) in the industrial sector. Differentiating equation (33) with respect to "n" we get the following expression, which will be positive when an increase in the number of industrial firms (keeping constant the number of enterprises) increases the effectiveness of monetary policy. Having defined (with reference to the multiplier 33): [- Sfl ] Al ={ a22 a33 + [a12 a23 (-1)] + sBM + [a13 (ý(")-l) a32 ]- [a13 a22 (-1)] - [a12 ('(")-1) a33] + r Sfl 'j } (35 - a23 a32 L- SBM J we have: dAl dü " Jl -" Al do do d(dk/dBM)/dn = (36 J2 We can see in the appendix that dJ /dn < 0. If, as we assumed earlier dk/dBM > 0, then we also have Al > 0. The expression /dn) Al, the -(dJ " second addendum at the numerator of expression 36, is therefore positive. For what concerns the first addendum, we have (see appendix for algebraic details): /\/ -+ S Sfl (' ](a22 - [- Sfl ] drB dAl/dn= a33-a23 832)+ (a22a33-a23a32)+ On L-SBM SrB SBM do 54 + /\ + / \ /\ +i 84(") drL +-- + (a13 a32-a12 a33) + (6a13/6n)(4(")-1)a32 + örL do --+ /\/\/ ++++-- (6a12/6n) (8a13/6n) - (6a12/8n)(ýt(")-1)a33 - a23 + a22 (37 Therefore, defining: S [- 6f lJ ED = (1/ a) (a22 a33-a23 a32) >0 (38 Sn 6BM ö [- öfl ] drB EB = (1/J. ) (a22 a33-a23 a32) <0 (39 örB ÖBM do 8-1)( ") drL EL = (1/J) (a13 a32-a12 a33) >0 (41 SrL do ES=(1/11)[(6a13/6n)(, D(")-1)x32 - (6a12/6n)(4ý(")-1)a33-(6a12/6n) a23+ ] + (6a13/6n) a22 (42 gJ. _ (1/J_ý, 2) [(dl- /dn)"Al] >0 (43 we will have: d(dk/dBM)/dn = ED + EB + EL + ES + EJ_ý, (44. The sign of ES (from definition 42) is uncertain, while the signs of ED, EB, EL, EZ, are unambiguously determined. Equation 44 tells us that the effect of an exogenous increase in the number of industrial firms (decrease in the degree of concentration) on the monetary policy multiplier dk/dBM can be decomposed into five parts. EB and EL can be intuitively thought of as the effect on dk/dBM 55 induced through a variation in the equilibrium level of rL and rB, originated by an increase in n. ED can be thought of as a direct effect on the multiplier dk/dBM of a variation in n, comparable to the one considered in the simplified case of section 2. Eft is the effect of a variation of n on the determinant of matrix "A", on the left hand side of system 27. ES is, like ELF, a "structural" effect of a variation in n. I call them "structural" because they can be intuitively thought of as variations (due to a change in "n") in the system "sensitivity" to monetary policy and market perturbations. For example, ES contains the effects, induced by a variation of "n", on the marginal productivity of capital, due to the fact that a variation in "n", by changing the equlibrium level of "k", changes the point of the production function chosen by each enterprise. ES also contains the effects of a variation of "n" on the elasticity of the slope of the credit supply function, since, a different point on the function S(. ) is "picked up" by the firms, again due to a modification of "n". EJ contains analogous effects, taking the form of modifications in the determinant of matrix "A" on the left hand side of the equality of system (27. In other words, the direct effect on dk/dBM of an increase in n, also brings some market perturbations (i. e. effects on the equilibrium values of the interest rates and other variables, as well as variations in the "sensitivity" of the system to market signals). As we can see from equations and definitions 37,38,39,40, 41,42, and 43, ED, EJL and EL are positive, ED is negative, and ES has an uncertain sign. Obviously, an increase in the degree of competition (reduction in the degree of concentration) in the 56 industrial sector, will increase the effectiveness of monetary policy on the level of investments if the positive effects prevail over the negative ones. However, we can conclude that, in general, a change in the oligopsonistic power of industrial firms on credit market should affect the effectiveness of the monetary policy, except in the very particular and extreme case where the negative elements of multiplier (44) exactly compensate the positive ones. 4. Conclusions. The models presented in this paper described an economy characterized by industrial firms oligopsonistic in the market for credit. Some effects of an exogenous variation of the market power of the industrial firms have been analyzed using two models. In the first one, the behaviour of the monetary sector is captured by a supply function of bank credit to the industrial firms, which depends positively on the (unique) interest rate, and on a generic parameter 0, whose value is bigger the more expansionary the monetary policy. In this simplified case, an increase in the degree of competition in the industrial sector (decrease in the degree of concentration) increases the effectiveness of monetary policy. This happens because an increase in the degree of competition, reducing the existing spread between marginal productivity of capital and interest rate on the credit market, creates an expansionary effect. The model of section 3 attempts to generalize the analysis of section 2, and describes the behaviour of the banking sector in accordance with the portfolio allocation theory. Having introduced 57 into the model a banking system and- other agents allocating their wealth according to the portfolio allocation theory, weakens the results of section two. However, a first result is that, in general, an exogenous change in the market power of industrial firms on credit market indeed affects the transmission mechanism of monetary policy. The direct effect (defined as "ED") of an exogenous increase in "n" (i. e. an increase in the degree of competition) is positive and analogous to the one considered in the simplified case. Some ambiguity appears in considering the sign of the indirect effects. The indirect effect induced through the equilibrium value of the interest rate on bank loans rL (defined as EL) is still positive. The indirect effect (again of an increase in the number of industrial firms) on the monetary policy multiplier 37 induced through the equilibrium level of rB is negative, and the effect on the system "sensitiveness" to market signals, (mainly determined by the reciprocal interactions of several agents allocating their wealth among several assets, and represented by ES + EJ.,, ) has an uncertain sign. However, in this model, a situation of irrelevance of the degree of concentration in the industrial sector on the transmission mechanism of monetary policy can only happen in the very extreme and particular case where the opposite effects affecting the monetary policy multiplier exactly compensate one another. 58 Bibliography of Chapter 2 Baltensperger, E. [1980] "Alternative Approaches to the Theory of the Banking Firm", Journal of Monetary Economics, vol. 6, pp. 1- 37. Banca d'Italia, [1989] "Atti del Seminario Ristrutturazione Economica e Finanziaria delle Imprese", Roma, 1988. Benassy, J. P., [1989] "Microeconomic Foundations and Properties of a Macroeconomic Model with Imperfect Competition", Sept. 1989, Paris, No. 8927, CEPREMAP. Deshmukh, S., Greenbaum, S. I., Kanatas, G. [1983], "Bank Forward Lending in Alternative Funding Environment", The Journal of Finance, vol. 38, pp. 873-886. Diamond, D., [1984], "Financial Intermediation and Delegated Monitoring", Review of Economic Studies, vol. 51, pp. 393-414. Duca, J. V., Vanhoose, D. D., [1990] "Loan Commitments and Optimal Monetary Policy", Journal of Money, Credit and Banking, vol. 22, pp. 178-194. Fischer, S. [1983] "A Framework for Monetary and Banking Analysis", Economic Journal, vol. 93, pp. 1-16. Gertler, M., Gilchrist, S. [1991] "Monetary Policy, Business Cycles and the Behaviour of Small Manufacturing Firms", IMF Seminar Paper, November. Hörngren, L. [1985] "Regulatory Monetary Policy and Uncontrolled Financial Intermediaries". Journal of Money, Credit and Banking, vol 17, pp. 203-219. Modigliani, F., Papademos, L., [1980], "The Structure of Financial Markets and the Monetary Mechanism", in Controlling Monetary Aggregates III, Conference Series no. 23, Federal Reserve Bank of Boston. Modigliani, F., Papademos, L. [1987], "Money, Credit and the Monetary Mechanism", in De Cecco, M., Fitoussi, J. P., Monetary Theory and Economic Institutions, London, Mac Millan Press. Santomero, A. M., [1984] "Modeling the Banking Firm", Journal of Money, Credit and Banking, vol 16, pp. 576-602. Tobin, J., Brainard, W. C., [1963] "Financial Intermediaries and the Effectiveness of Monetary Control", American Economic Review, vol. 53, pp. 383-400. VanHoose, D. D., [1985], Bank Market Structure and Monetary Control", Journal of Money, Credit and Banking, 17, No. 3. 59 Appendix System 27 is obtained by differentiating at the equilibrium functions fl, f2, and f3. We consider then linear approximations of the functions fl, f2, and f3. This implies that the second derivatives of the behavioural functions are null. We also assume that the generical derivative Sfi/SrLSrB is null. öfl 6 r S(-) (1/n)" _ 6BM 6BML SS(")/örL 84, (") SBb SLd ]- St( .) [ý(")-1]"L "BM -- [BM(41, (")-1)-Bb-Ld] 1 SrL SrL SrL SrL ">0 n rSý(") SB SL 2 "I "BM -- LSrL SrL SrL 6f2 -= ['(") - 1] >0 SBM 8f3 -_ -1 6BM Sfl 1 S(. ) <0 Sn n2 SS(. )/SrL Sfl = a(a-1)Bka-2 = all <0 Sk hfl 1 S(-) -1 -S /SrL ; SrL n 8S(. )/6rL which can be rearranged as follows: Sfl [1 E -1 -1 SrL n where E= [öS(. )/örL][rL/S(. )] is the credit supply elasticity with respect to the interest rate; and rL E= [82S(")/6rL2][ SS(. )/SrL 60 is the elasticity of the slope of the same function; but since we are considering linear approximations of the functions at the equilibrium, we will have E=0, hence: öfl 1 _ -1 -= a12 <0 SrL n For what concerns the remaining signs, we will have: Sf1 6S(. )/rB =-= a13 >0 örB öS(. )/SrL n 6f2 =N a21 >0 6k 6f2 SS(. )/SrL = a22 <0 SrL öf2 SS(. )/SrB = a23 >0 SrB Sf3 = a31 = 0. 6k 6f3 = SBb/SrL + 6Bp/8rL = a32 <0 örL Sf3 = SBb/örB + SBP/örB = a33 >0 SrB Therefore, the sign pattern of matrix A on the left hand side of system 27 is the following: -- + +- + o- + the determinant of the matrix A is: --+++--++-+- 3= (all a22 a33) +(a13 a21 a32)-(a12 a21 a33)-(all a23 a32) Since we assumed that in each demand or supply function the derivatives with respect to their own interest rate are bigger, in absolute value, than the cross derivatives, we have: 61 Ja22J > ja231 ; 1a331 > Ja321 and 1a121 > ja131; which implies fam. > 0. In addition we have: 8a12/6n = n-2 > 0; SS(")/örg 6a13/6n = n-2" < 0; SS(")/SrL d(cD(")-1)/dn = (S-t(")/örL)"(drL/dn) 8f1 6f1 d_ oft 1_S [_ 16 ] drB ö_ L 6f1 I drL + + do SBM Sn öBM SrB SBM do SrL SBM do where: s afl -1)-n-2(-, 5f /b' BM) < 0. 8r 8BM S 6fl]=[(ý? /8rL)BM+(_oBb/8rL)+(_ýLd/ýrL)]_2[(8/8rL)(8S( 6888SrL . )/ýrL)+ L- 6BM + (öý/örL)(-öS(")/SrL)l (1/n)= 0 S "fl SrB L- 8BM J=E(5ý/6rL)BM+(-8Bb/8rL)+(-6Ld/8rL)]-2 (6Ld/6rB)3) . {(64/6rL)[(6Bb/8rB) + >0 At this point we are enabled to calculate dAl/dn and d, L1./dn, which appears at the numerator of equation 36. (-1)n-2[(4(")-1)(8S(")/6rL) + (8(P/6rL)"(S("))] dAl/dn = [(öý/SrL)"BM + (-SB /örL)+(-öL /SrL)]2 62 1 S'' "[(8Bb/8rB)+(8Ld/6rB)] (drB/dn) n SrL "(a22 a33-a23 a32) + [(S, D/SrL)"BM + (-8Bb/8rL)+(-6Ld/8rL)l2 SP drL 8S(")/6rB "(a22 a33-a23 832) + (a13 a32-812 a33)+(n-2) (-t-1) " SrL do SS(")/örL 8S(")/6rg "a32 + (-1) (n-2)x23 + (n-2) a22 -ýY1ý2ýýý-1)x33. SS(")/SrL Sall dk 6a13 8a12 dL-\, /dn = (a22 a33-a23 a32)+ (a21 a32)- (a21 a33)" 6k do on On Since we assumed that in each demand or supply function the derivatives with respect to their own interest rate are bigger, in absolute value, than the cross derivatives, we have: 6a12_ Sa13 > and ßa331 > ja321 on ön which implies dL\, /dn < 0. 63 CHAPTER 3 CAPITAL MARKETS' SOPHISTICATION AND BANK LENDING: AN INTERPRETATION IN THE SPIRIT OF 0. WILLIAMSON (1985) AND AN EMPIRICAL ANALYSIS *I am very grateful to Keith Cowling, Norman Ireland, Giovanni Amisano and David Vanhoose for helpful comments. I am obviously the only one responsible for any mistakes that might be found and for the views expressed here. 64 CAPITAL MARKETS' SOPHISTICATION AND BANK LENDING: AN INTERPRETATION IN THE SPIRIT OF O. WILLIAMSON (1985) AND AN EMPIRICAL ANALYSIS. 1. Introduction. In this chapter it is argued that the distinction made in some "institutional" analyses between "securitized" and "non securitized" financial sectors is more than a purely theoretical classification and may indeed carry some macroeconomic implications, empirically detectable, for what concerns the behaviour of the credit supply to industry. For this purpose, some empirical analysis will be implemented to show how the different behaviour of bank credit to industry in two different institutional contexts could be interpreted in the light of Williamson's [1985] contractual relations framework. While in most countries the most relevant source of finance is provided by profits retentions, it is well known that stock and bond markets are relevant sources of financial funds for the industrial firms in the U. S. and in the U. K. only. In continental Europe, the source of external finance is predominantly provided by the banking system. A distinction is made between the countries whose financial sectors are "intermediaries oriented" and the ones whose financial sectors are "securitized" is made by, among others, Rybczynski [1984]. While Japan is often quoted as a typical example of non- securitized financial system, the distincion between securitized and non-securitized financial systems, might be informative for a few European countries. Gardener [1991], following Rybczynsky's classification, argues that the United Kingdom is, at the moment, in 65 the phase of -"securitization", France and Germany are approaching it, while Italy is only entering an intermediate stage of gradual development of financial markets. The reason for such an evolution could lie in the increasing efficiency of markets and in their increasing capacity for risk bearing. -Similar analyses have been made by Frankel and Montgomery [1991] and Mayer [1992]. The latter also points out that in the German "non-securitized" financial sector (unlike in the British "securitized" one) hostile takeovers are a very rare phenomenon. This fact, due to a higher dispersion of share ownership in the "securitized" financial sectors, suggests that in non securitized financial sectors the connection between market for shares and market for control could be less direct and less straightforward. Outside Europe, the most significant example of "securitized" financial sector is the United States, while Japan is an often- quoted example of a financial system strongly oriented towards intermediaries. The impressive economic growth of Japan and Germany in the last decades shows that no value judgement-may be associated to the concept of "securitization". In the "intermediary-oriented" financial systems, a direct control on bank credit by the government authorities is, theoretically speaking, feasible (although not necessarily advisable), and constitutes an historically relevant experience. In what follows, Williamson's [1985] contract relations framework will be adopted in order to interpret the existence and development of securitized and non-securitized financial systems. A description of Williamson's approach and its implications are in 2. explained section Sections 3 and 4 contain an empirical 66 analysis of the implications of the two different financial systems. In particular, it will be argued that the phenomenon of securitization can make the demand for bank credit by industrial firms more unstable than the supply. Empirical evidence in favour of this last point will be provided by an analysis based on British and German data. Section 5 contains the conclusions drawn from the analysis. 2 An interpretation of securitization based on 0. Williamson's [1985] contractual relations framework. Williamson's [1985] approach, based on the relevance of transaction costs, suggests an interpretation of the behaviour of economic agents in terms of contractual relations: the relevance of transaction costs (that Williamson defines - by quoting Arrow - as the "costs of running the economic system") is also implied by those interpretations of the Arrow-Debreu model defining commodities not only by physical, spatial, and time characteristics, but also by those elements of environmental uncertainty referred to as the "state of the world". Economies on transaction costs can be implemented by assigning transactions to governance structures chosen among different institutional alternatives: the "classical market contracting" at one extreme, a centralized hierarchical organization at the opposite side, and mixed models of firm and market organization in between. In this context a relevant role is played by the assumption of "bounded rationality" and the analysis of agents' opportunistic behaviour. The relevance of bounded rationality contrasts with the 67 traditional approach which suppresses the role of institutions in favour of the interpretations of firms as "production functions", or "black boxes". In Williamson's words: "Confronted with the realities of bounded rationality, the costs of planning, adapting, and monitoring transactions need expressly to be considered. ... Transactions that are subject to ex post opportunism will benefit if appropriate safeguards can be devised ex ante. " (Williamson [1985], pp. 46-48). In such a context, "asset specificity" is a conceptual tool that contains many informationally relevant elements (such as the frequency of the transaction) for the decision process described by the bounded rationality approach. In a context of unbounded rationality, contracts would determine a world of planning. On the other hand, in a world with transaction costs, "the imperative that is this: organizational emerges ... organize transactions so as to economize on bounded rationality while simultaneously safeguarding them against the hazards of opportunism. Such a statement supports a different and larger conception of the economic problem than does the imperative 'Maximize profits'!. " (Williamson, cit. p. 32). In what follows the theoretical contracting schema presented in Williamson [1985] will be applied to bank credit. Instead of the "supply for a commodity", the "supply for financial funds" will be "general used, instead of the purpose technology", a "spot market for financial funds" will be used, and instead of Williamson's 68 "transaction specific asset", the credit supplied by the bank to the firm be specific will usedll . The contracting schema could be represented as follows: A * Pi k=0 / B P2 \ s=0/ k>0 \/ P2 > P3 s>0 \ P3 C FIGURE 1. Referring to figure 1, if one defines "k" as a measure of transaction-specific sunk cost for collecting information about borrowers' riskiness (the correspondent of Williamson's transaction specific assets), financial funds may be supplied on a bonds market, or intermediated and supplied by the banks. Banks can be thought of as agencies specialized in performing economies of scale in 11 Such a definition shows some similarity with Okun's [1981] distinction between "auction markets" and "customer markets". According to Okun, in the latter, the kind of contract ties together particular sellers to particular buyers (in our case, particular financial intermediaries to particular firms), creating a sort of long-run relationship where the quantities (in our case the financial flows) or the prices (in our case the interest rates) might be fixed in the short run. 69 collecting information. In Williamson's terminology, the first case corresponds to the general purpose technology, the second one to the special purpose technology. Since parties have an interest in creating safeguards to protect investments in transactions, "s" can be defined as the magnitude of those safeguards which could be assimilated to the collaterals in banks' loans. Following Williamson [1985] again, one can assume that suppliers are risk neutral, willing to supply under either kind of transaction, and accept any safeguard condition provided that expected breakeven results can be obtained. In the absence of transaction-specific sunk costs (k=0), node A is reached; this corresponds to a breakeven price (interest rate) pi. The node B is reached in the presence of transaction-specific sunk costs (k>O) without safeguard (in our case without collateral, i. e. s=0). In the contract corresponding to node C we have transaction-specific sunk costs in the presence of safeguard (k>O, s>O). k, s, and p are determined simultaneously by a contract, and obviously influence each other. Transactions performed under each of those regimes can take place at the same time. In particular, in the real world, we see banks supplying credit with or without collateral (obviously at different breakeven levels of interest rates) when they lend different amounts of money to different categories of borrowers, and, above all, we see the coexistence of bonds markets and intermediated credit. We can further imagine that, since in node A transactions take place without asset specificity,. such a situation may be reached when the state of nature is such that the number of relatively low- risk agents is high enough to trigger a supply of financial funds 70 under a general purpose contract. This may happen if the governance institution is able to prevent forms of moral hazard and opportunistic behaviour that would undermine the feasibility of general purpose contracts. In other word, the existence, efficiency, size and macroeconomic relevance of stock markets might depend on the effectiveness of governance structure (and, of course, on all the other features influencing the number of relatively low-risk borrowers and agents, which might have to do with the degree of economic development). Securitized financial systems could be thought of as institutional contexts where the magnitude of the financial flows negotiated under the kind of contracts associated to node A is macroeconomically relevant compared to those associated to nodes B and C. We can further think of nodes B and C as different "hierarchical" contractual forms of banks loans. Williamson's that "... transaction located B to observation at node ... are apt (and) be unstable contractually ...... may revert to node A ... or be relocated to node C" (Williamson [1985], p. 34), could be interpreted as a description of phenomena relating to credit rationing (i. e. unwillingness to supply credit to those agents unable to provide collateral) or other situations where the supply of bank credit to industries is affected by the willingness of banks to lend, and their subjective valuation on the riskiness of lenders. Such subjective valuations might be affected by the business cycle, as in the Bernanke and Blinder [1988] model. 71 2.1 Securitization: empirical relevance for bank credit and some macroeconomic implications. If, as Fama [1985] suggests, bank credit is a non substitutable source of finance for the firms penalized on bonds and shares markets by phenomena of asymmetric information and agency costs, then in a "securitized" financial system banks would conceivably face two different kinds of customers: firms non penalized on capital markets, for which the different sources of funds are substitutable, and firms penalized on capital markets. The firms of the first type (usually big corporations with a "strong" reputation), due to the high substitutability among the different sources of funds, would probably express a more unstable demand for banking credit than the firms of the second type. Fama's [1985] analysis, is obviously not inconsistent with the financial economics literature which describes the dividend decision by the firm's controlling group as a problem of signaling (Leland and Pyle [1977]): in the presence of information asymmetry, if the dividend policy can be used as a "signal" (on the quality of the firm's investment) in order to reduce the transaction costs, smaller firms (with lower market power and lower profit flow) might not be able to send the"signal" that would enable them to reduce the transaction costs. For this reason, they may be more dependent on the supply of credit by agencies specialized in monitoring and in economies of scale in collecting information (i. e. banks). For the same reason, in securitized financial systems, where the spot market for financial funds is empirically relevant, substitution between bank credit and securities is likely to be an empirical phenomenon too. 72 To the extent that the bank credit to large corporations represents a big portion of the total bank credit to industry, the demand for bank credit expressed by "strong" firms would contribute to determine an empirically observable stock level of bank credit to industry consistent with the behaviour of a supply function rather than a demand function. In fact, if the demand function for bank credit is more unstable than the supply (due to the high substitutability between bonds, shares and bank credit for the industrial firms), then, looking at the equilibrium stock of bank credit (and estimate with a monoequational model) a supply function is observed, rather than the usual demand function. This contrasts with the approach followed by most empirical works, which estimate a demand function with the assumption of partial adjustment. On the other hand, in a non securitized financial sector there is no reason to expect a demand for bank credit to industry more unstable than the supply. Therefore, a demand function for bank credit to industry could be identified in this last case. All these points deserve an intuitive explanation. Let us consider a supply of bank credit (in our case to the industrial sector) analogous to the one employed in the theoretical part of Bernanke and Blinder [1988], i. e. : +- Ls = 1i(rL, rß)"D(1-t) (1 where LS = banks' credit supply; D= banks' deposits; -c = banks' required reserves coefficient; rL = interest rate on banks' loans; 73 rg = interest rate on bonds; Let us assume that the following function ++- D= D(Y, P, rD) (2 describes the deposits supplied by the banks to the public. In this function (where Y is defined as the national income, P the price level, and rD as the interest rate on banks' deposits) the deposits supplied by the banks are assumed to depend positively on the aggregate income, on the price level, and negatively on their cost, i. e. the interest rate rD. Let us further assume that a stable functional link exists between the two liquid assets traded on the spot markets: bank deposits and'bonds. Let us define this functional link as follows rD=rD(rB); (3 the behaviour of such functional link may be determined by the term structure of the interest rate. The above-mentioned case, where the demand for bank credit is more unstable than the supply for bank credit, may be intuitively described with the help of the following figure (assuming, only for the sake of this graphic, that the competition between the different banks determine a fixed spread between rL and rD, so that rL may be expressed as a function of rD, and, figure 1 may be drawn, for simplicity, with one interest rate only): Ls = supply of bank credit to the industrial firms; Ld = demand for bank credit by the industrial firms 74 FIGURE 2 rL LS Ls, Ed Figure 2 shows (although in a simplistic way) that if, in a monoequational context, the demand function is more unstable than the supply, the simple observation of the stock values of the bank credit to the industry would identify a supply rather than a demand function. In fact, when the credit demand shifts from Ldl to Ld2 and Ld3, the three observable equilibrium stocks of credit lie in the same supply function, which is then identified. It is argued here that in a "securitized" financial system, the bank credit to industry is highly substitutable by the recourse to the "spot" market. Therefore, the demand for bank credit by industrial firms might shift when conjunctural causes affect the transaction and monitoring costs by making the spot market comparatively less *or more convenient. Substitutability between bank credit and non- intermediated credit is possible both in a securitized and in a non- securitized financial sector. If, according to the institutional analysis, substitutability is expected to be empirically relevant in securitized financial systems and not in the non-securitized ones, a for bank supply function credit to industry can be estimated, in the 75 former and not in the latter, using macroeconomic data. It must be stressed that observable data correspond to the equilibrium levels of bank credit to industry. In other words, since they correspond to actual negotiations, they might be thought of as equilibrium values between demand and supply: they are, per se both demand and supply values, and it is the relatively larger instability of one of the two functions'that allows the identification of the other. A further element which may contribute to increase the instability of the demand for bank credit is the diffusion of "loan commitment" contracts. In such contracts the bank commits itself to provide the customer with an overdraft availability, at the interest rate prevailing on the market at the moment of the actual utilization of the loan. The "loan commitment" contract is more flexible for the firm, which may decide not to use the credit availability (reducing, in that case its financial costs). On the other hand, the loan commitment brings higher opportunity costs in case the interest rate prevailing at the moment of the actual use of the loan availability is higher than the interest rate on long term loans at an initial moment, when a normal spot contract could be stipulated. In such a framework, a loan commitment contract is convenient for the firm if the expectations of variability in (and aggregate demand then the possible risk of undertaking unnecessary financial costs) are higher than the expectations of increase in the interest rate on loans. Duca and Vanhoose [1990] increasing diffusion show that an of loan commitment contracts could alter the macroeconomic impact of, monetary policy. A large diffusion of loan commitment contracts would-also carry higher instability of for bank industry. the demand credit to Duca and Vanhoose also point 76 out that the empirical evidence for the United States shows that loan commitment contracts have a lower bankrupt ratio than other loan contracts. This seems to suggest that the banks tend to stipulate loan commitment contracts with less risky firms. Also, quoting the "Federal Reserve Bulletin" ("Terms of Lending at Commercial Banks: Survey of loans made", vol. 74, September 1988), they remark that the share of loan contracts stipulated under the form of "loan commitment" tend to increase with the size of the single loan contracts: "... This suggest that large firms, which probably pose less credit risk and are thus' less likely to be rationed, tend to borrow under commitments to a greater extent than smaller, and probably less secure, firms" (Duca and Vanhoose, [1990], p. 180). it must be said, however, that the actual diffusion of such a kind of contract is not easily detectable, because its main element is the continuity in the relationship between bank and industrial firm. This element can be determined also by extensions or renegotiations of the terms of the contracts, which may not be explicitly observable and may constitute one of the main features of the relationship between bank and customer in the non-securitized financial systems. 3. An empirical analysis of some macroeconomic implications of securitization: the United Kingdom and Germany. To perform a comparative analysis of the behaviour of bank credit to industry in the two different kinds of financial systems, the United Kingdom and Germany, have been taken into consideration, 77 during the period between the 1974 oil shock and the German Unification. Such a choice is due to the fact that the magnitude of the two economies is comparable, and, as Frankel and Montgomery [1991] point out, many legal issues and structural features (such as the degree of concentration in the banking sector) are very similar in spite of the two countries having very different financial systems. Thus, by comparing the behaviour of bank credit in the two countries, it might be possible to "isolate" (at least to some extent) the effects of the two different financial systems. In Frankel and Montgomery [1991], a comparison of several different legal issues between Germany, U. K., U. S. and Japan shows that "regulation of British and German banks follows a universal bank model, under which banks are permitted to engage in a wide range of financial activities, including all insurance and securities activities. The main difference between the British and the German versions of the universal bank is that British banks usually conduct their securities business through subsidiaries, while German banks conduct their business directly" (Frankel and Montgomery, (1991), p. 273). The regulation issues analyzed by Frankel and Montgomery are: the principal regulators of commercial banks, geographic and regulatory banking restrictions, the scope of permissible activities (such as securities, insurance, industrial investments) capital requirements, deposit protection scheme, and reserve requirements. No relevant differences have been found, in this regard, between the British and the German banking systems, while some relevant differences are pointed out in a few institutional features which are, incidentally, at the core of the attention in most studies on securitization, namely the customer relationship and the bankruptcy procedures. For 78 what concerns the former point, in Germany (like Japan), banks are able to establish "very close ties" with industrial firms, whilst this happens only very rarely in the U. K. and in the U. S. For what concerns the bankruptcy procedures, while the British (and American) law heavily penalizes banks that have close relationships with customers and imposes greater losses on the banks than on other creditors, not only are banks less penalized in Germany than in Britain, in case of customers' distress, but "often take responsibility for organizing creditor coalitions for financially troubled firms. a bank's behaviour in such a workout may be disciplined by its interest in establishing and maintaining a reputation as a structurer and arranger of successful firms' finance" (Frankel and Montgomery (1991), p. 288). For this reason, informal bankruptcy arrangements are in Germany more frequent than the informal ones, and this may help to explain bankruptcies why the number of corporate has greatly increased in the U. K. (and in the U. S. ), while it does not show any clear upward trend for Germany (and Japan). Frankel and Montgomery [1991] also report some empirical data concerning the trend of real assets of the largest banks (dramatically increasing for Germany, more stable for Britain, between 1970 and 1989), and the comparison between funds raised through securities and bank loans. The data reported in Frankel and Montgomery [1991] clearly show that by comparing all the 1965 1989 quinquennia between and in the U. K., any increase in loans is aggregate bank associated with a decrease in funds raised through securities, and viceversa. Such phenomenon (which, been curiously, has not pointed out by Frankel and Montgomery) only affects the U. K. and the U. S. (i. e. the two securitized financial 79 systems) and seems to be consistent with the fact that (as argued in this paper) the securitized financial systems are characterized by higher instability in the demand for bank credit to industry than non-securitized financial systems. This might also be consistent with Mayer's [1992] observation of the fact that bank credit is the most relevant "anti-cyclical" source of financial funds for the industrial firms. For what concerns the degree of concentration in the banking sector, the data reported by Frankel and Montgomery show that the concentration of bank assets in the five largest banks is very similar between the U. K. and Germany, while the German banking system seems to be slightly more concentrated if one looks at the ten largest banks. Data on the degree of concentration in the banking sector are also reported by Gardener [1991], on the basis of OECD data, and the review "The Banker" (for what concerns the concentration ratios). Gardener's [1991] data are reported in the following tables. Degree of market concentration and size of the banking systems for Germany and the United Kingdom in 1988. concentration share of the market. total assets deposits 5 banks 3 banks 5 banks 3 banks Germany 31.2 21.2 30.5 19.1 United Kingdom 32.6 26.5 30.3 21.6 number of size banks of the banking sector Assets (billions of $) Germany 4465 1465.0 United Kingdom 661 1337.8 80 According to the concentration ratios calculated with 3 or 5 banks, the British banking system seems to be only slightly more concentrated than the German one. On the other hand, if one calculates the ratio between the size of the banking sector (measured by the assets in billions of dollars) and the number of banks operating in the market, it can be seen that the average size of the banks is larger in the United Kingdom than in Germany. The British and the German banking sectors seem, then, to differ in all the features (such as customer relations, bankruptcies procedures and substitutability between bank credit and securities) connected with the phenomenon of securitization, while they do not show any relevant difference for what concerns the regulatory issues and the market structure. In addition, both economies are in a stage of advanced industrialization, their magnitude is comparable (at least until the German Unification), and they are highly integrated (which could lead us to assume that they are subjected to the same sources of disturbances, at a macro level). For these reasons, a comparative analysis on the behaviour of bank credit to industry should be able to provide information on the institutional differences of the two financial systems by isolating (at least to some extent) their macroeconomic implications. For what concerns Germany, bank credit will be modelled following the standard approach (demand function containing an interest rate representative of the cost of borrowing, and another interest rate containing information on the money market conditions) because, being the German financial system non-securitized, between substitutability securities and bank credit is not On the hand, empirically relevant. other bank credit to industry in 81 the United Kingdom will be described by a supply function, analogous to the one contained in the theoretical part of the paper by Bernanke and Blinder [1988]. This because in the securitized British financial system substitutability between bank credit and securities is expected to be empirically relevant. The empirical specification derived from such a function contains an interest rate representative for bank assets and another for the banks liabilities. If these two interest rates have very similar coefficients, they may capture (if jointly considered) the information for the interest rate spread between banks' liabilities and banks' assets, which could be interpreted either as a mark-up pricing mechanism, or (if one does not assume imperfect competition in the banking sector) as a proxy for the margin of intermediation necessary in order to cover the administrative, costs of the bank. The analytical form of the supply for bank credit to industry is log linear like the following: Ct =ap + al"ln(Yt) + a2"ln(Pt) + a3"rL - a5"rD (4 This equation may be obtained by substituting equations 2 and 3 into equation 1, and approximating the resulting equation with a Cobb-Douglas, exponential in the interest rates. A more complex dynamic structure of equation 4 will be obtained by following the general-to-specific approach. In this case, equation 4 would be form considered in the more general bl(L)ln(Yt) b2(L)ln(Pt) b3(L)rL Ct =bp + + + - b5(L)rD (4' 82 where b2(L), b3(L), b4(L), b5(L) are lag polynomials, on which a series of zero restrictions have to be tested, as explained in the next section. 3.1 A Brief description of the econometric methodology. The purpose of the following analysis is to show how the behaviour of the bank credit to the industrial firms is affected by such an institutional feature as the relevance of the stock market. To do so, the behaviour of some relevant credit aggregates for Germany and United Kingdom will be compared and contrasted. For the sake of completeness, the demand for money will also be taken into account. The specifications of the different equations have been obtained following the "general-to-specific" (Hendry [1985], Harvey [1989]) methodology, starting from a general unrestricted specification containing four lags. Simulation studies have shown that this seems to be an appropriate dynamic structure to start with in order to capture the dynamic properties of the models, while several studies in the 1980's (for example Hendry [1988] and Muscatelli [1988]) have shown that Feedback mechanisms (like the one at the basis of the general-to-specific approach) yield better econometric performances that the "forward-looking" ones (for instance Cuthbertson [1985]). The appendix contains the tables with the data, the estimations and the results. All the estimations have been implemented with 3.0. In "Microfit", version the tests, the level of confidence of 95 has been used unless otherwise specified. . 83 3.2 General-to-specific and partial adjustment. The partial adjustment mechanism can be consistent with an optimizing behaviour of economic agents. It can be assumed that individuals face costs due to the fact that they are holding a quantity of money different from the one of long run equilibrium; and other adjustment costs. Considering a quadratic cost function, the decision of the agents can be described as follows: min C= a(Mt - M*)2 + b(Mt - Mt-1 )2 where C is the total cost, Ma financial asset. The asterisk stands for "desired" level; "a" is the cost of holding a level of asset different from the equilibrium one, "b" represents the adjustment cost. Assuming that the second order conditions for optimization are satisfied, the first order conditions are the following: 6c = 2a(Mt - M) + 2b(Mt - Mt-1 )=0 SMt ab_ Mt =M+ Mt-1 = 6M + (1 - 6)Mt-1 a+ba+b with 6= a/(a+b). In the case of the money demand, we have M=k+ aYt - ßRt + ut (where "Y" is the income, "R" an interest rate representative of the conditions of the money market, "k" a constant, "u" a stochastic disturbance). then the estimable equation describing the short run demand for money is: (1-6)Mt_l Mt = Ak + AaYt - AßRt + + but Mt = const + g1Yt + 92Rt + g3Mt-1 + Vt where const = 6k; gl=6a; g2=8R; 93 = (1-8); Vt=eut. 84 A similar procedure and analytical function would be obtained for the demand for credit. In the estimates unadjusted data have been employed, because, as Wallis [1974] points out, the use of seasonally adjusted data could induce distortions in the estimations, apart from the very particular case where the same lag operator applies for the dependent variable and the regressors. The estimates have been performed with the method of ordinary least squares, but a test of exogeneity on income has been implemented in the final specification of the functions of demand for money and credit, in order to detect simultaneity between income and the dependent variable considered. For this purpose the Hausman exogeneity test has been implemented. This test is composed of two phases: in the first phase the variable subject to exogeneity test must be regressed on an instrument. The residuals of this regression must be included, in a second phase, in the original regression, containing the variable subject to exogeneity test. If the coefficient referred to the residuals does not seem to be significant (according to the statistics), then the variable subject to test can be considered exogenous with respect to the independent variable. 4. The Empirical Results In what follows I will comment briefly on the estimates for each country and the relative economic implications. 85 4.1 The United Kingdom. In the case of the United Kingdom, three credit aggregates have been considered: the bank credit to industry, the bank credit to sectors other than industry, and the total bank credit (given by the sum between the bank credit to industry and the bank credit to the other sectors). The qualitative behaviour of the bank credit to industry differs a great deal from the other two credit aggregates. However, the information referred to the bank credit to industry seems to be more interesting because it identifies a well-defined category of credit users. For the reason mentioned earlier, a supply function for bank credit to industry has been estimated for the U. K., instead of a demand function. This supply function contains the interest rate on seven days notice deposit account with London clearing banks, (here defined RLCB, which can be regarded as a leading interest rate on banks' deposits market) and the interest rate on banks' overdrafts (ROV). The former is an indicator for the interest rates on banks' liabilities, the latter is obviously an interest rate on banks' assets. The spread between these two interest rates could represent a proxy for the "gross margin of intermediation" existing in the banking system. If we accept such'an interpretation, in an hypothetical estimation of the supply function of bank credit to industry, the coefficient referred to RLCB should be negative, while the one referred to ROV should be positive, and, if one regards the difference "ROV-RLCB" as a proxy for the banks' for margin of profit, or the margin of intermediation, they should have very close absolute value. We could imagine, as a first approximation, a partial adjustment mechanism of the kind: St - St-1 = e(S*t - St-l) 86 with S*t = S(ROVt, RLCBt, Pt, Yt) where the asterisk indicates "desired value", St is the supply of bank credit to industrial firms at time t, Pt the level of prices, Yt the real output. The lagged dependent variable mechanism could be justified with the same kind of argumentations which justify a partial adjustment mechanism for a demand function for a financial asset. Preliminary analyses have shown that, even if the data employed are non seasonally adjusted, seasonal dummies were not significant in any credit aggregate "general" unrestricted specification (unlike the demand for money equation, as we will see later). Therefore seasonal dummies were not included in the general unrestricted models employed for the "general-to-specific" analysis. In the estimates referring to the total credit the interest rates employed were analogous to the ones used by Bernanke and Blinder [1988]: the interest rate on short run treasury bills (TREBIRA) and the interest rate on bank overdrafts (ROV). In some preliminary analyses, ROV turned out to be more significant than the prime rate. Tables 1 and 2 in the appendix show the general unrestricted bank model for the total credit to the economy. There seems to be a structural break at the end of 1988 (i. e. 1988 QIV), corresponding to the period where the Bank of England definitively dropped Ml as an intermediate target for monetary policy: in fact, the model 1975 estimated over the sample period QII to 1991 QIV largely fails the test for normality of the residuals (and marginally fails the test of serial correlation); the same model estimated over the 1975 QII to 1988 sample period QIV largely fails the predictive 87 failure test calculated on the data of 1989,1990 and 1991, while it largely passes all of the other diagnostic tests. The general unrestricted model for the bank credit to other sectors than industry (tables 3 and 4) yields very similar results'and also shows as well a structural break at the end of 1988: the model estimated over the sample period' 1975 QII to 1988 QIV fails the predictive failure, while the model estimated over the sample 1975 QII to 1991 QIV largely fails the test for normality of the residuals. Both bank credit to other sectors than industry and the total bank credit supports a partial adjustment mechanism with a lagged dependent variable (tables 5 and 6 respectively). Such models, determined by imposing some parameters restrictions on the general models of tables 4 and 2, are estimated over the sample period 1975 QII to 1988 QIV, in order to run the predictive failure test, for the years after the Bank of England had dismissed Ml as an intermediated target monetary policy. LBACROIN (table 5) fails the predictive failure and the Chow test, and does not fail all of the other diagnostic tests. The same equation estimated over the sample period 1975 QII - 1991 QIV (table 7) largely fails the test for normality of the residuals. On the other hand, the "restricted" model for the total bank credit (LBACRTO, table 6, over the sample period 1975 QII QIV) all the diagnostic tests, including (although _ 1988 passes of failure largely marginally) the predictive and the Chow test. It fails the test for normality of the residuals when estimated over the sample period 1975 QII - 1991 QIV (table 8). These results bank to suggest that the credit other sectors than industry and the total bank credit seem to be quite sensitive to changes in the Eargeting, do monetary policy ; and not seem to be very stable. The 88 results are quite different for what concerns the bank credit to industry (LBACREIN) which seem to be much more stable than the other two credit aggregates. For the reason illustrated earlier, LBACREIN has been estimated as a supply function. Tables 9 and 10 show the general unrestricted model with four lags estimated over the sample period 1975 QII - 1991 QIV and 1975 QII - 1988 QIV respectively. The model of table 9 marginally fails the test on the normality of residuals at the level of confidence of 0.95 (while it does-not, fail it at the level of confidence of 0.99), and the model of table 10 yields satisfactory results in all of the diagnostic tests, including the predictive failure one. However the low number of degrees of freedom of the estimates of table 9 weakens the reliability of the tests. In both sample periods LBACREIN supports the parameters restrictions necessary to obtain a "partial adjustment" model with a lagged dependent variable (tables 13 and 14 respectively). Since it is somehow "unusual" to estimate a supply function for bank credit to industry in a monoequational framework, the procedure implemented in order to obtain the final dynamic specification have been shown. Tables 11 and 12 respectively show the variable deletion tests for the restrictions allowing to obtain the restricted model of table 13 from the general model of table 9, and the restricted model of table 14 from the unrestricted one of table 10. In both sample periods, the "restricted" models-yield very in the diagnostic satisfactory results all of tests, including the the Chow predictive failure ones and test. This seems to suggest that the "supply function" of bank credit to industry seem to be than the "demand function" much more stable for bank credit to other 89 sectors than industry and the "demand function" for total bank credit. One of the possible reasons for such a result is the fact that the bank credit to industry refers to a category of borrowers which is much more homogeneous than for the other two aggregates. In particular, the total bank credit is a highly heterogeneous aggregate, which might be affected by a much larger set of disturbances. Tables 15,16 and 17 refer to the Hausman test of exogeneity calculated for the variable LRUKYDS (log of the British real GDP) in the "restricted" specifications of LBACREIN, LBACROIN and LBACRTO respectively. In all of the three cases the null hypothesis of exogeneity of the variable LRUKYDS is not rejected at the level of confidence of 0.95. 4.2 Germany. If one had to give an example of structural break, one of the best would probably be that of Germany in 1989-1990. Since testing for structural stability on the post-unification data would yield too obvious results, the regressions have been run on the sample period 1975 QI - 1989 QIII (i. e. until the announcement of the opening of the East German frontiers, which could rationally be interpreted as the first step of unification), while the last quarter of 1989 and the first one of 1990 have been employed to run a meaningful predictive failure test, since the formal process of unification took place gradually after the elections of 1990. The bank credit to manufacturing sector has been estimated as a demand function with partial adjustment, since in this case the 90 for assumptions by Fama 1985 and the empirical evidence substitutability between security and bank credit do not apply, given that the German financial system is not "securitized". A set of interest rates analogous to those employed by Bernanke and Blinder [1988] has been used (namely the interest rate on banks' credit on current account and the interest rate on treasury bills). Observing the graphic with the path of the dependent variable (LBACREMA in Figure 3), it can clearly be seen that in 1980, term 3 there is a structural break, probably determined jointly by the effects of the implementation of the European Monetary System and by the recession which took place that year. In order to apply the general-to specific methodology, a general unrestricted model with four lags has been estimated using the data from 1975 QII to 1989 QIII. However, (as one can see from tables 18 and 19 for the estimates without seasonal dummies and with seasonal dummies respectively) the general unrestricted model largely fails the diagnostic test on the normality of residuals, due to the structural break of 1980. Therefore a model with partial adjustment has been estimated over the sample period from 1980-QIV to 1989-QIII, without testing the coefficients restrictions on a general unrestricted because model with lagged variables, there would not have been freedom enough degrees of to make the variable deletion tests 20 reliable. Table shows the estimates with OLS over the sample 1989 period 1975 QII - QIII, and table 21 shows the same model estimated over the reduced sample period 1975-QII to 1980-QIII, both which largely fails the predictive failure and the Chow test for structural stability, proving that there is indeed a structural 411. ILn Cr C-71 . --Q 141-:b cc cm 110 UD cc cri cn r-I X> -4 cc co 0.1-Si rcD cc cc C`«.5 u: s N r: b 91 break in 1980 QIV. The coefficient of the implicit deflator of GNP (LIPGNP) has a wrong sign, but the variable is not significant. Tables 22,23 and 24 refer to the Hausman test of exogeneity for the variable LRGGNP (log of the real German GNP). In particular, table 22 shows the regression of LRGGNP on the instruments employed for the test (a constant and the lagged values of LRGGNP, the implicit deflator of GNP LIPGNP, the German treasury bill rate GTRBR, the log of Ml LNM1, and the log of the bank credit to manufacturing sector LBACREMA). The residuals HREBCMA of such regression have been included in the regression for LBACREMA in order to implement the second step of the test, shown in table 23 and 24 (the regression without seasonal dummies and with seasonal dummies respectively). In both cases the model fails the Hausman test for exogeneity for the variable LRGGNP. Therefore new estimates have been implemented using the method of instrumental variables. In particular, tables 25 and 26 show the instrumental variable estimation of the equation without seasonal dummies and with seasonal dummies respectively. Again, in both the regressions of table 25 and 26, the coefficient for the implicit deflator of the GNP has a wrong sign, but it is not significant. The estimates with the dummies have been shown only for the sake of completeness, since the dummies do not seem to be significant in this case. For what concerns Germany, the demand for bank credit by manufacturing industries does not seem to be very stable. It is very sensitive to changes in the business cycle, like the structural break in 1980 seems to show. Therefore the estimates for LBACREMA do not seem to perform well enough to provide reliable information on demand the use of the for bank credit to industry as a possible 92 intermediate target for monetary policy. However, this may be due to the fact that the observable values of LBACREMA are just equilibrium stocks, and could be thought of as linear combinations of demand and supply functions. In this case LBACREMA has been interpreted in the former way, because it has been argued that in a non securitized financial sector there is not high substitutability between bank credit and securities, or bonds. As a consequence, in a monoequational framework one can follow the predominant approach in the existing empirical literature, and estimate a demand function for bank credit. In order to run a "counterfactual" experiment, it is shown, in tables 27 and 28, that estimating a supply function for banks' credit analogous to the one estimated for the U. K. (i. e. containing a set of interest rates which could be interpreted as a proxy for the banks' gross margin of intermediation) does not yield any sensible result. The results of table 27 (estimation with instrumental variables) do not make much sense, while the equation shown in table 28 (estimate with OLS) looks more like a demand function where the interest rate RTDEP acts as a proxy for GTRBR. Therefore, the prediction that securitized financial sectors would make the demand for bank credit by industry more unstable than the supply seems to be confirmed by the data. 4.3 A quick comparison with the demand for money The behaviour of the various credit aggregates have been compared to the one of the money demand, in order to obtain further information by comparing the stability and statistic performances of the two functions. The main purpose of the estimates commented in 93 this section is to detect the presence of structural breaks (possibly different from the ones affecting the credit stocks). Like in Bernanke and Blinder [1988], the money aggregate employed both for the U. K. and Germany is Ml. The estimates for the money demand in the U. K. have been implemented only over the sample period where Ml has been used as an intermediate target of monetary policy, i. e. 1975 QII - 1988 QIV. This is also the period relevant for the comparative analysis on credit aggregates, since the comparison of the data after 1989 are obviously biased by the beginning of the process of unification in Germany. In addition, the data later than 1988 for Ml (i. e. after Ml has been definitively abandoned as an intermediate target for monetary policy) have not been reported by any OECD statistical publication, and only a few quarters later disappeared even from the CSO financial statistics. Table 29 shows the general unrestricted model for the money demand, which gives satisfactory results for all of the diagnostic tests. The dummy variables have been included because they appeared to be more significant than in the previous pieces of analysis. Table 30 shows the variable deletion test which leads to the "parsimonious" specification with partial adjustment shown in table 31. The signs and magnitudes of the parameters seem to be consistent with the standard theory. The coefficient referred to the implicit deflator of GDP is not significant, and the model marginally fails (at the level of confidence of 95%) the diagnostic test for serial 32 correlation. Table shows the Hausman test for exogeneity calculated for the variable LRUKYDS (log of the real GDP), which turns out to be exogenous at the level of confidence of 95% and 90% (marginally) . 94 In the context of Germany, the equation for the demand for money seems to perform much better than the demand for bank credit to manufacturing sector. Tables 33 to 36 show the general-to- specific analysis. Table 33 shows the general unrestricted model, which yields satisfactory results for all of the diagnostic tests, excepting, obviously, the predictive failure test calculated over the last quarter of 1989 and 1990, due to the drastic changes in the monetary policy regime, immediately after the announcement of the end of Berlin's wall and the beginning of the process of unification. The seasonal dummies have been omitted because they turned out to be largely non significant in a preliminary analysis. Obviously this does not mean that seasonality is disregarded: it is assumed, instead, that the seasonal effects are captured by the dynamic structure of the model. The variable deletion test in table 34 shows that the general unrestricted model does not support a partial adjustment mechanism, and requires a more complex dynamic structure, which has been determined through the variable deletion test of table 35 by keeping the most statistically significant regressor in the general unrestricted model. The final specification is shown in table 36. The diagnostic tests yield satisfactory results (excepting again the predictive failure test calculated over the period after the announcement of the unification), and the coefficients of the various regressors seem to have signs and magnitudes correspondent to the predictions of economic theory, apart from the more complex dynamic structure. In particular, it can be seen that the real GNP affects Ml with a delay of three quarters, the level of Ml shows some hysteresis, and seems to react to some 95 weighted time-difference in the price level, rather than to the price level itself. 5. Conclusions. The empirical analysis of the previous sections show that the mechanisms characterizing the credit market are heavily influenced by institutional features (such as whether a financial system is market-oriented or bank-oriented). Also, the theoretical framework commonly employed to interpret the behaviour of money stock does not necessarily describe appropriately in any institutional context the behaviour of another liquid asset such as bank credit to industry, which may be affected by the specificity of the assets, determined by the interactions between industrial firms and financial intermediaries. For the sake of the present analysis, the most relevant credit aggregate is credit to industrial firms. The simplifying assumption of the traditional IS-LM model, which only envisages a distinction between money and generical "bonds", and implicitly aggregates any form of credit (i. e. bank credit and securities) in a unique financial sector, might turn out to be unreliable, due to the role of asset specificity and endogenous contractual response that characterizes the behaviour of the financial sector. The role of endogenous contractual response becomes empirically relevant in the context of securitized financial systems, because of the substitutability between bank credit and securities. 96 The institutional factors taken into account in this analysis are the distinction between "securitized" financial systems (like the British one), "intermediaries-oriented" financial systems (like the German one). Such a distinction is significant, especially in connection with Fama's [1985] assumptions, according to which some classes of enterprises, facing agency costs in the financial markets due to informational asymmetry phenomena, are heavily dependent on bank credit. In such a context banks could face two kinds of customers (in the industrial sector): "strong" enterprises which can easily substitute bank credit by issuing assets on the capital markets, and firms penalized by access costs in the financial markets. The demand for bank credit expressed by the "strong" firms could be more unstable than the supply, due to the high substitutability between direct emission of financial market assets and recourse to the bank credit. Both the situations contribute to create the conditions which allow for the estimation (within a monoequational model) of a supply' function of bank credit, rather than a demand function, starting from the observation of credit stock. 'Such a situation seems to apply in the "securitized" financial sectors, in our case in the United Kingdom, where the following supply function of bank credit to industry has been estimated. LBACREIN=. 79068+. 19881 LRUKYDS+. 64950 LNIP+. 032855 ROV-. 028242 RLCB+ (1.5756) (2.3964) (9.6660) (7.5303) (-6.8997) +. 49464 LBACREIN_1 (10.2420) R2=. 99796; the numbers in brackets refer to the t-statistics. On the other hand, for the total bank credit (where it is not possible to picture a particular class of agents having homogeneous 97 characteristics) a demand function has been estimated, according to the approach most commonly followed in empirical works, and similarly to what has been done by Bernanke and Blinder [1988]. The fact that a supply function for bank credit to industry-was identified yields some evidence (perhaps not very strong) in favour of the theoretical interpretation suggested in section 2,2.1 and 3. The empirical literature often implicitly assumes that the stock of credit is "demand determined". This is an interpretation once more based on the implicit assumption of macroeconomic irrelevance of the decisions of the banking sector concerning the allocation of credit, like in the IS-LM model. In other words the relevant behavioural relation would be a demand function and the existing stock would differ from the one desired by agents only because of adjustment costs, lags in correcting expectations, or modifications in the equilibrium levels of the relevant variables. Such a specification derives often from the application to credit demand of the specifications commonly employed to describe money demand, and implicitly assumes that the supply of banks' credit adjusts to the shocks and modifications originating on the (credit) demand side of the market. Following a "Poole-like" analysis (in the spirit of Bernanke and Blinder [1988]), one could argue that in a securitized financial system, if the monetary policy has to rely on the most stable behaviourial relation, then, to the extent that a policy target may be represented by the volume of activity of the industrial sector, some proper intermediate target could be chosen among the variables appearing in the supply function of bank credit to the industrial sector. A monetary policy based on a strict control of money 98 aggregates may be less effective in a securitized sector, where substitutability between different, sources of financial funds and endogenous contractual response could determine other forms of "virtual liquidity". Finally, the present empirical analysis seems to testify to the relevance of institutional features in determining the behaviour of financial assets, as argued by many post-keynesian economists, like Davidson [1982]. 99 Bibliography of Chapter 3 Arestis, P. [1988] "Post-Keynesian Theory of Money, Credit and Finance", in "Post-Keynesian Monetary Economics", ed. by P. Arestis, 1988 Edward Elgar, Aldershot (U. K). Angelopi, F. [1984] "Il mercato degli impieghi bancari in Italia: Un'analisi econometrica (1974-1982)" Banca d'Italia Banca d'Italia, Atti del Seminario [1988] "Ristrutturazione Economica e Finanziaria delle Imprese", Roma. Baltensperger, E. [1980] "Alternative Approaches to the Theory of the Banking Firm", Journal of Monetary Economics 6,1-37. Bera, A. K., Jarque, C. M. [1981], "An Efficient Large-Sample Test for Normality of Observations and Regression Residuals, Working Paper in Econometrics no. 40, Canberra, Australian National University. Bernanke, B. e Blinder, A. [1988] "Is it Money or Credit, or Both, or Neither? Credit, Money, and Aggregate Demand" American Economic Review: Papers and Proceedings, May, Princeton University, New Jersey Bernanke, B., Mishkin, F. [1992] "Central Bank Behaviour and the Strategy of Monetary Policy: Observations from Six Industrialized Countries", NBER Working Paper no. 4082, May. Conigliani, C., [1990], "La Concentrazione Bancaria in Italia", Bologna, Il Mulino Conti, V. (1991) "Concorrenza e redditivitä nell'industria bancaria: un confronto internazionale" in Banca d'Italia: temi di discussione del Servizio Studi 149, febbraio Corradi, V., Galeotti, M., Rovelli, R., [1988], "I1 Sistema Bancario e ii Meccanismo di Trasmissione", in "Banca e Mercato", ed. by Cesarini, F., Grillo, M., Monti, M., Onado, M., Bologna, Il Mulino. Cuthbertson, K. (1988) "The Demand for Ml: A Forward-Looking Buffer Stock model", oxford Economic Papers 40,110-131 Davidson, J. E. H, Hendry, D. F., Srba, F. e Yeo, S. (1978) "Econometric Modelling of the Aggregate Time-Series Relationship between Consumers' Expenditure and Income in the United Kingdom", Economic Journal 88,661-692 Davidson, P. (1982), "International Money and the Real Word", MacMillan, London Diamond, D. (1984) "Financial Intermediation and Delegated Monitoring" Review of Economic Studies 51, July, 393-414 Dufour, J. M., [1980], Dummy Variables and Predictive Tests for 100 Structural Changes, Economic Letters, vol. 6, pp. 241-247. Fama, E. (1985) "What's Different about Banks? ", Journal of Monetary Economics, 15, January, 29-40 Frankel, A., B., Montgomery, J., D. (1991) "Financial Structure: an International Perspective", Brookings Papers on Economic Activity, n. 1, Washington D. C., Brookings Institutions. Gardener, E. P. M., (1991) "Nuove strategie per le banche europee", Banca Impresa, Societä Anno X, 1, aprile; 19-39; Il Mulino, Bologna Gardener, E. P. M. and Molyneux, P., (1990) Changes in Western European Banking; Unwin Hyman, London Giannini, C., Papi, L., Prati, A. (1991) "Politica di offerta e Riallocazione del Credito Bancario negli Anni Ottanta", Temi di Discussione, Banca d'Italia, febbraio Godfrey, L. G., [1978a], "Testing Against General Autoregressive and Moving Average Error Models When the Regressors Include Lagged Dependent Variables", Econometrica, vol. 46, pp. 1293-1301. Godfrey, L. G., [1978b], "Testing for Higher Order Serial Correlation in Regression Equations When the Regressors Include Lagged Dependent Variables", Econometrica, vol. 46, pp. 1303-1310. Gorton, G., Kahn, J. (1992), "The Design of Bank Loan Contracts, Collateral and Renegotiation", University of Rochester, Rochester Center for Economic Research, Working Paper no. 327. Harvey, A. C. (1989) The Econometric Analysis of Time Series; Philip Allan, Deddington Hendry, D. F. (1979) "Predictive Failure and Econometric Modelling in Macroeconomics: The Transactions demand for Money" in Ormerod, P. (ed. ), Economic Modelling; Heinemann, London Hendry, D. F. (1985) "Monetary Economic Myth and Economic Reality", oxford Review of Economic Policy 1,72-84 Hendry, D. F. (1988) "The Encompassing Implications of Feedback versus Feedforward Mechanisms in Econometrics" Oxford Economic Papers 40,132-149 Hendry, D. F. Mizon, G. E. (1978) "Serial Correlation as a Convenient Simplification, not a Nuisance: A Comment on a Study of the Demand for Money by the Bank of England" Economic Journal 88 549-563 Jarque, C. M., Bera, A. K., [1980], "Efficient Tests for Normality, Homoscedasticity and Serial Independence of Regression Residuals", Economic Letters, vol. 6, pp. 255-259. Leland, H. and Pyle, D., (1977) "Information Asymmetries, Financial Structure and Financial Intermediation", The Journal of 101 Finance, May. Mayer, C. (1990), "Financial systems, Corporate Finance and Economic development" in Asymmetric Information, Corporate Finance and Investment, ed. by G. Hubbard, Chicago, University of Chicago Press. Mayer, C. (1993) "Corporate Finance", forthcoming in The New Palgrave Dictionary of Money and Finance, London, MacMillan Press Minsky, H. P. (1986), "The Evolution of Financial Institutions and the Performance of the Economy", Journal of Economic Issues, June. Muscatelli, V. A., [1989], "A Comparison of the 'Rational Expectations' and 'General-to-Specific Approaches to Modelling the Demand for Ml", Oxford Bulletin of Economics and Statistics, 51, pp. 353-375. Ramsey, J. B., [1969], "Tests for Specification Errors in Classical Linear Least Squares Analysis", Journal of the Royal Statistical Society, series B, vol. 31, pp. 350-371. Ramsey, J. B., [1970], "Models, Specification Error and Inference: a Discussion of Some Problems in Econometric Methodology" Bulletin of the Oxford Institute of Economics and Statistics, vol. 32, pp. 301-318. Rybczynsky, T., (1984) "The Financial System in Transition" National Westminster Bank Quarterly Review November Salkever, D. S., [1976], The Use of Dummy Variables to Compute Prediction Errors and Confidence Intervals", Journal of Econometrics, vol. 4, pp. 393-400. Santomero, A. M., (1984) "Modelling the Banking Firm", Journal of Money, Credit and Banking 16. Wallis, K. F., [1974], "Seasonal Adjustment and Relations Between Variables" Journal of the American Statistical Association, 69, pp. 18-31. Williamson, 0. (1985) "The Economic Institutions of Capitalism", New York, The Free Press. 102 APPENDIX. List of the variables employed in the regressions. United Kingdom: BACREIN = bank credit to the industry; BACROIN = bank credit to other sectors than industry; BACRTO = BACREIN + BACROIN; IP = implicit price deflator of GDP; LBACREIN = log(BACREIN); LBACROIN = log(BACROIN); LBACRTO = log(BACRTO); LNIP = log(IP); LNMl = log(M1); LRUKYDS = log(UKGDP/IP); RLCB = interest rate on deposit account at seven days notice with London clearing banks; ROV = interest rate on banks' overdraft; TREBIRA'= interest rate on treasury bills; UKYDS = UK GDP at market prices; UHAUS = residuals for the Hausman test of exogeneity of LRUKYDS; Germany: BACREMA = bank credit to the manufacturing sector GTRBR = German treasury bill rate; HREBCMA = residuals for the Hausman test for exogeneity for LRGGNP in the equation for LABCREMA; GGNP = GNP at current prices; IPGNP = implicit price deflator of the GNP; LIPGNP = log(IPGNP); LNM1 = log of Ml LBACREMA = log(BACREMA); LRGGNP = log(GGNP/IPGNP); RCRCA = interest rate on bank credit on current account; CONST = intercept term; Si, S2, S3 = seasonal dummies for the first, second and third terms repsectively. Source of data: DATASTREAM SERVICES at the University of Warwick, on the basis of seasonally unadjusted OECD data. Diagnostic Tests The diagnostic test performed in the tables that follows are those provided by the package MICROFIT 3.0. In particular, the main references are the following. 103 The diagnostic test for serial correlation is the one suggested by Godfrey [1978a], [1978b]. The diagnostic test for the functional form is Ramsey's RESET test (Ramsey [1969], [1970]). The diagnostic test for the normality of the regression (Jarque residuals is the Jarque-Berg test and Bera [1980], Bera and Jarque [1981]). The following pages contain the tables with the estimates and the diagnostic tests. The diagnostic test for heteroscedasticity is based on the auxiliary regression e2t = const + ayt the the the fitted where et are residuals of regression and yt dependent the the values of the variable. auxiliary regression gives LM and F-test for the null hypothesis HO: a=0. The diagnostic test for predictive failure is the one suggested by Salkever [1976] and Dufour [1980]. S I03. a A note on Hausman Test The variable UHAUS corresponds to the residuals of the regression of LRUKYDS on a constant term, LNM1(-1), LNIP(-1), TREBIRA(-l) ROV(-1). It has been employed to run the Hausman test of in exogeneity of LRUKYDS the equations of the different credit for United in demand for aggregates considered the Kingdom and the money. The variable HREBCMA corresponds to the residuals of the regression of LRGGNP on a constant term, LNM1(-1), GTRBR(-1), LIPGNP(-1), LRGGNP(-1), LBACREMA(-1). It has been employed to run the Hausman test of exogeneity of LRGGNP in the demand for money and in the demand for bank credit by the manufacturing sector for Germany. I f TABLE I Ordinary Least Squares Estimation Dependent variable is LBACRTO 67 observations used for estimation from 7502 to 9104 Regressor Coefficient Standard Error T-Ratio[Frob] CONST -2.0358 . 83005 -2.4526[. 018] 11573 LRUKYDS -. . 22205 -. 52117[. 605] TREBIRA 0039656 0078004 50838[. 614] . . . ROV 0029105 0078261 . . . 37189[. 712] 11723 LNIP . . 45402 25820[. 798] 87233 . LBACRTO(-1) . . 14047" 6.2099[. 000] 083276 LRUKYDS(-1) -. . 17858 -. 46633[. 643] TREBIRA(-1) 0055581 0087271 . . . 63687[. 528] 012758 0085051 ROV(-i) -. . -1.5000[. 141] LNIF(-1) -1.0562 . 71822 -1.4706[. 149] 070153 LBACRTO(-2) . . 19410 36143[. 720] 19811 . LRUKYDS(-2) . . 17952 1.1036[. 276] 013295 0089514 TREBIRA(-2) . . 1.4852[. 145] 011686 0085800 ROV(-2) -. . -1.3620[. 180] 1.1411 74117 LNIP(-2) . 1.5396[. 131] 14279 LBACRTO(-3) -. . 19065 -. 74897[. 458] LRUKYDS(-3) 039115 17631 22185[. . . . 826] TREBIRA(-3) 0022162 0092368 . . . 23993[. 812] 0010068 ROV(-3) -. . 0086834 -. 11594[. 908] 72609 LNIP(-3) -. . 74357 -. 97650[. 334] LBACRTO(-4) 10437 12671 . . . 82371[. 415] LRUKYDS(-4) 39537 . . 24929 1.5860[. 120] TREBIRA(-4) 0097853 0082237 . . 1.1899[. 241] 0095220 ROV(-4) -. . 0076518 -1.2444[. 220] 59539 LNIP(-4) . . 44998 1.3232[. 193] R-Squared . 99950 F-statistic F(24, 42) 3524.6[. 000] 99922 R-Bar-Squared . S. E. of Regression 024995 Squares . Residual Sum of . 026240 Mean of Dependent Variable 11.5937 Variable 89506 S. D. of Dependent . Maximum of Log-lik elihood 167.7445 DW-statistic 1.9561 Diagnostic Tests * Test Statistics * LM Version * F Version * A: Serial Correlation *CHI-SO( 4)= 10.4741[. 033]*F( 4, 38)= . 1.7603[. 157]8 * B: Functional Form SCHI-SO( 1)= 060794[. 805]*F( . 1, 41)= . 037236[. 848]* * C: Normality *CHI-SO( 2)= 72.7916[. 000)* Not applicable 4 * D: Heteroscedasticity *CHI-SO( 1)= 1.7691[. 183]*F( 1, 65)= 1.7629[. 189]3 test A: Lagrange multiplier of residual serial correlatio n RESET test the B: Ramsey's using square of the fitted val ues test C: Erased on a of skewness and kurtosis of residuals Eased the D: on regression of squared residuals on squared fitted values TABLE 2 Ordinary Least Squares Estimation ***********#*******#*## ***#**####******#******#*******************************: Dependent variable is LBACRTO 55 observations used f or estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] CONST -1.2184 . 76809 -1.5862[. 123] 24495 LRUKYDS -. . 23711 -1.0331C. 310] 0068020 0074765 90978[. TREBIRA . . . 370] 0011040 0075895 ROV . . . 14547[. 885] 53695 42697 LNIP . . 1.2576C"2181 82726 LBACRTO(-1) . . 15212 5.4380[. 000] 078376 18113 668] LRUKYDS(-1) -. . -. 43270[. 0026383 755] TREBIRA(-1) -. . 0083816 -. 31477[. 0077774 ROV(-1) -. . 0076742 -1.0134[. 319] LNIP(-1) -1.4062 . 68857 -2.0422[. 050] 014378 21016 LBACRTO(-2) . . . 068414[. 946] 11944 17277 69134[. LRUKYDS(-2) . . . 495] 024531 TREBIRA(-2) . . 0089570 2.7388[. 010] 019632 ROV(-2) -. . 0080627 -2.4349[. 021] 95055 LNIP(-2) . . 72384 1.3132[. 199] 041719 19306 21609[. 830] LBACRTO(-3) -. . -. 0065664 LRUKYDS(-3) -. . 17341 -. 037867[. 970] 0048277 TREBIRA(-3) -. . 010074 -. 47920[. 635] 0042089 0093340 ROV(-3) . . . 450920.6551 LNIP(-3) -1.1133 . 71012 -1.5677[. 127] 12326 13338 LBACRTO(-4) . . . 92415C. 363] 49234 25226 LRUKYDS(-4) . . 1.9517[. 060] 010556 TREBIRA(-4) . . 0093633 1.1274C. 2693 0086290 ROV(-4) -. . 0090824 -. 95008[. 350] 1.1006 LNIP(-4) . 44528 2.4718[. 019] 99951 R-Squared . F-statistic F(24, 30) 2567.8[. 000] 99912 S. E. R-Eiar-Squared . of Regression . 021051 Residual Sum of Squares . 013294 Mean of Dependent Variable 11.3063 S. D. of Dependent Variable . 71132 Maximum of Log-lik elihood 150.9714 DW-statistic 1.7663 Diagnostic Tests * Test Statistics * LM Version * F Version Correlation *CHI-SCI( 4)= 4.8242[. 306]*F( * A: Serial 4, 26)= . 62495[. 649]4 Form *CHI-SO( 1)= 1.7476[. 186]*F( * B: Functional 1, 29)= . 95171[. 33774 * C: Normality SCHI-SCE( 2)= 1.5298C. 465]* Not applicable A * D: Heteroscedasticity SCHI-SO( 1)= 1.1547[. 283]*F( 1, 53)= 1.1366[. 291]* * E: Predictive Failure *CHI-S0( 12)= 29.2123[. 004]*F( 12, 30)= 2.4344[. 024]4 A: Lagrange multiplier test of residual serial correlatio n B: Ramsey's RESET test using the square of the fitted val ues C: Sased on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on square d fitted values E: A test of adequacy of predictions (Chow's second test) TABLE 3 Ordinary Least Squares Estimation Dependent variable is LBACROIN 67 observations used for estimation from 7502 to 9104 Regressor Coefficient Standard Error T-Ratio[Prob] CONST -. 79924 . 68418 -. 90393[. 371] 22933 28561[. 777] LRUKYDS -. 065499 . -. 0080818 69975[. 488] TREBIRA . 0056553 . . 0080968 048332[. 962] ROV . 3913E-3 . . 092510[. 927] LNIP . 044714 . 48334 . LBACROIN(-i) 1.1450 . 15281 7.4926[. 000] 11110 18976 561] LRUKYDS(-1) -. . -. 58545[. 50858[. 614] TREBIRA(ri) . 0046143 . 0090731 . 133] ROV(-1) -. 013353 . 0087181 -1.5317[. 57895[. 566] LNIP(-1) -. 43357 . 74889 -. 017760 2367i 074399[. 941] LBACROIN(-2) -. _ . -. 14276 19286 74019[. 463] LRUKYDS(-2) . . . TREBIRA(-2) . 012706 . 0093915 1.35290.1833 234] ROV(-2) -. 010728 . 0088941 -1.2062[. 55450 74932 740000.4633 LNIP(-2) . . . 30347 24124 215] LBACROIN(-3) -. . -1.2580[. LRUKYDS(-3) -. 039560 . 18767 -. 21080[. 834] 3915E-3 0094871 041269[. 967] TREDIRA(-3) . . . 0089616 29640[. 768] ROV(-3) . 0026562 . . LNIP(-3) -. 42116 . 73911 -. 56982[. 572] 12930 15556 411] LDACROIN(-4) . . . 83121[. 23999 25652 355] LRUKYDS(-4) . . . 93556[. 0049479 0084412 58616[. 561] TREBIRA(-4) . . . ROV(-4) -. 0059477 . 0078452 -. 75613[. 453] 31894 70765[. 483] LNIP(-4) . . 45071 . R-Squared . 99958 F-statistic F(24, 42) 4130.5[. 000] 99933 S. E. R-Dar-Squared . of Regression . 025871 028112 Residual Sum of Squares . Mean of Dependent Variable 11.2562 S. D. of Dependent Variable 1.0029 Maximum of Log-lik elihood 165.4357 DW-statistic 2.0887 Diagnostic Tests * Test Statistics * LM Version * F Version * A: Serial Correlation *CHI-SO( 4)= 6.5790[. 160]*F( 4, 38)= 1.0344[. 402]1 * S: Functional Form *CHI-SQ( 1)= 2.0358[. 1547*F( 1, 41)= 1.2848[. 264]8 * C: Normality *CHI-SO( 2)= 87.0336[. 000]* Not applicable A * D: Heteroscedasticity *CHI-SO( 1)= 3". 0557[. 080]*F( 1, 65)= 3.1061t. 083)* A: Lagrange multiplie r test of resid ual serial correlation B: Ramsey's RESET tes t using the squ are of the fitted val ues C: Based on a test of skewness and k urtosis of residuals D: Based on the regre ssion of square d residuals on squared fitted values TABLE 4 Ordinary Least Squares Estimation Dependent variable is LBACROIN 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] 15181 80987 18745[. 853] CONST . . . 22406 334] LRUKYDS -. . 22820 -. 98188[. 0066164 0071439 362] TREBIRA . . . 92617[. 0024822 0073719 33671[. 739] ROV . . . 44920 LNIP . . 42707 1.0518[. 301] 1.1725 LBACROIN(-1) . 16748 7.0007[. 000] 18873 LRUKYDS(-1) -. . 18510, -1.0196[. 316] 623] TREBIRA(-1) -. 0040258 . 0081026 -. 49685[. 0094500 ROV(-1) -. . 0073266 -1.2898["207] 67732 115] LNIP(-1) -1.1004 . -1.6246[. 057313 27470 20864[. 836] LBACROIN(-2) . . . 11694 LRUKYDS(-2) . . 18464 . 63266[. 532] 022463 TREBIRA(-2) . . 0089046 2.5227[. 017] 018938 ROV(-2) -. . 0080074 -2.3651[. 025] 48539 69523 69818[. 490] LNIP(-2) . . . 38855 LBACROIN(-3) -. . 26118 -1.4677[. 147] 12429 LRUKYDS(-3) -. . 17922 -. 69354[. 493] 0074315 TREBIRA(-3) -. . 0096489 -. 77019[. 447] 012685 ROV(-3) . . 0091406 1.3878[. 175] 58278 LNIP(-3) -. . 65771 -. 88608[. 383] 15244 17208 88588[. 383] LBACROIN(-4) . . . 41259 LRUKYDS(-4) . . 24899 1.6570[. 108] 0026362 TREBIRA(-4) -. . 0091496 -. 28813[. 775] 0022472 0091198 807] ROV(-4) . . . 24641[. 75093 LNIP(-4) . . 41699 1.8008[. 082] R-Squared . 99964 F-statistic F(24, 30) 3511.5[. 000] 99936 S. E. Regression 020388 R-Bar-Squared . of . 012470 Residual Sum of Squares . Mean of Dependent Variable 10.9378 S. D. of Dependent Variable . 80556 Maximum of Log-lik elihood 152.7325 DW-statistic 2.0335 Diagnostic Tests * Test Statistics * LM Version * F Version ********************************************************** **************##****a * A: Serial Correlation *CHI-SO( 4)= 9.0862[. 059]#F( 4, 26)= 1.2863[. 301]1 Eu Functional Form *CHI-SO( 1)= 70307[. 402]*F( 37551[. * . 1, 29)= . 545]9 * C: Normality *CHI-SO( 2)= 5.3441[. 069]* Not applicable I * D: Heteroscedasticity *CHI-SQ( 1)= 1.2486[. 264]*F( 1, 53)= 1.2312[. 272]X * E: Predictive Failure *CHI-SO( 12)= 37.6325[. 000]*F( 12, 30)= 3.1360[. 005]4 A: Lagrange multiplier test of residual serial correlatio n B: Ramsey's RESET test using the square of the fitted val ues C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) TABLE 5 Ordinary Least Squares Estimation Dependent variable is LBACROIN 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] 24058 CONST -. . 48144 -. 49971[. 620] 030167 090003 33518[. 739] LRUKYDS . . . 0093062 TREBIRA . . 0060473 1.5389[. 130] ROV -. 0080704 . 0059709 -1.3516[. 183] 091203 LNIP . . 032716 2.7877[. 008] 97143 LBACROIN(-1) . . 019598 49.5690[. 000] R-Squared . 99921 F-statistic F( 5,49) 12339.0[. 000] 99913 S. E. Regression R-Bar-Squared . of . 023823 Residual Sum of Squares . 027809 Mean of Dependent Variable 10.9378 S. D. of Dependent Variable . 60556 Maximum of Log-likelihood 130.6756 DW-statistic 1.5695 - Durbin's h-statistic 1.6137[. 107] Diagnostic Tests * Test Statistics * LM Version *F Version * # * ýI Correlation *CHI-SO( 4)= 2.7165[. 6c)6]*F( 4, 45)= 58453[. 675]4 * A. -Serial . * # * 8 Form *CHI-SQ( 1)= 458[. 431]*F( ]* * E4: Functional .6 1, 48)= . 54778[. 46. * * * 4 * C: Normality SCHI-SO( 2)= 3.1669[. 205]* Not applicable 8 * * 4C A SCHI-SO( 1)= 155o8[. * D: Heteroscedasticity . 694]*F( 1, 53)= . 14987[. 700]* * * * * K E: Predictive Failure *CHI-SO( 12)= 25.7471[. 012]*F( 12, 49)= 2.1456[. 031J* * * # I * F: Chow Test *CHI-SO( 6)= 12.9889[. 043]*F( 6, 55)= 2.1648[. 06O]* A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) F: Test of stability of the regression coefficients TABLE 6 Ordinary Least Squares Estimation Dependent variable is LBACRTO 55 observations used for estimation from 7502 to B804 Regressor Coefficient Standard Error T-Ratio[Prob] CONST -1.0193 . 50292 -2.0267[. 048] 19837 LRUKYDS . . 098910 2.0056[. 050] 0045923 0066625 TREBIRA . . . 66927[. 494] 0043673 ROV -. . 0065566 -. 66610[. 508] 13516 LNIP . . 040266 3.3567[. 002] 92411 LBACRTO(-1) . . 026786 34.5003[. 000] R-Squared . 99877 F-statistic F( 5,49) 7925.6[. 000] 99864 S. E. Regression R-Bar-Squared . of . 026242 Residual Sum of Squares . 033743 Mean of Dependent Variable 11.3063 71132 - S. D. of Dependent Variable . Maximum of Log-likelihood 125.3567 DW-statistic 1.6446 Durbin's h-statistic 1.3445[. 1793 Diagnostic Tests * Test Statistics * LM Version *F Version * * # : Serial Correlation *CHI-SO( 4)= 1.0685C. 899]*F( 45)= 22289C. 92431 * A: 4, . * * # * B: Functional Form *CHI-SQ( 1)= ;. 1858[. 0743*F( 1, 48)= 2.9512C. O9231 * * * * C: Normality *CHI-SO( 2)= 2.9399[. 230]* Not applicable % D: Heteroscedasticity *CHI-SQ( i)= 1.0496[. 376]*F( 1, 53)= 1.03111.315]4 * E: Predictive Failure SCHI-SO( 12)= 17.5673[. 129]*F( 12, 49)= 1.4639[. 171J * F: Chow Test *CHI-SO( 6)= 8.3930[. 211]*F( 6, 55)= 1.3988[. 2ä2]x A: Lagrange multiplier test of residual serial correlation E4: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: E$ased on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) F: Test of stability of the regression coefficients TABLE 7 Ordinary Least Squares Estimation Dependent variable is LBACROIN 67 observations used for estimation from 7502 to 9104 Regressor Coefficient Standard Error T-Ratio[Prob] CONST -. 71216 . 48714 -1.4619[. 149] 12476 089084 1.4004[. 166] LRUKYDS . . 0060654 TREBIRA 012572 . 2.0728[. 042] ROV -. 011654 . 0057443 -2.0289[. 047] 10470 LNIP . . 029404 3.5607[. 001] 95169 LBACROIN(-1) . . 015402 61.7914[. 000] R-Squared . 99936 F-statistic F( 5,61) 19077.9[. 000] 99931 S. E. Regression R-Bar-Squared . of . 026371 042422 Residual Sum of Squares . Mean of Dependent Variable 11.2562 S. D. of Dependent Variable 1.0029 Maximum of Log-likelihood 151.6514 DW-statistic 1.4756 Durbin's h-statistic 2.1635[. 031] Diagnostic. Tests * Test Statistics * LM Version *F Version ***# * A: Serial Correlation *CHI-SQ( 4)= 4.8465[. 303]*F( 4,57)= 1.1112C. 360]ß * BiFunctional Form *CHI-SQ( 1)= 5.3024[. 01]*F( 1,60)= 5.1566[. 027] * CiNormality *CHI-SO( 2)= 29.5368C. 000]ß Not applicable ýK D: Heteroscedasticity *CHI-SQ( 1)= 1.0968C. 295]*F( 1,65)= i. 4818[. 302]* A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Pased on the regression of squared residuals on squared fitted values TABLE ß Ordinary Least Squares Estimation Dependent variable is LBACRTO 67 observations used for estimation from 7502 to 9104 Regressor Coefficient Standard Error T-Ratio[Prob] 48056 009] CONST -1.2952 . -2.6952[. 24893 LRUKYDS . . 091091 2.7327[. 008] 0065996 0063085 TREBIRA . . 1.0461[. 300] ROV 0056548 0059626 94837[. 347] ti, -. . -. 12935 LNIP . . 032194 4.0179[. 000] 92014 LBACRTO(-1) . . 018629 49.3927C. 0003 99913 F-statistic F( 5,61) 14059.6[. 000] R-Squared .. 99906 S. E. Regression 027413 R-Dar-Squared . of . 045841 Residual Sum of Squares . Mean of Dependent Variable 11.5937 S. D. of Dependent Variable . 89506- Maximum of Log-likelihood, 149.0547 DW-statistic 1.5834 Durbin's h-statistic 1.72510.0853 Diagnostic Tests * Test Statistics * LM Version #F Version ***a Correlation *CHI-SO( 4)= 1.6737[. 795]*F( 36510[. 832J * A: Serial 4,57)= . Form *CHI-SQ( i)= 14853[. 700]*F( 716]' k B=Functional . 1,60)= . 13331[. ***> * C: Normality *CHI-SO( 2)= 10.7174[. 005] Not applicable ***3 Heteroscedasticity SCHI-SO( i)= Q486916[. 926]*F( 0084332[. 927]) * D: . 1,65)= . A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Dased on the regression of squared residuals on squared fitted values TABLE 9 Ordinary Least Squares Estimation Dependent variable is LBACREIN 65 observations used for estimation from 7502 to 9102 Regressor Coefficient Standard Error T-Ratio[Prob] 55726 96694 CONST . . . 57631[. 568] 14611 LRUKYDS -. . 26052 -. 56083[. 578] 1.6399 47509 LNIP . 3.4519[. 001] 031990 ROV . . 0075072 4.2612[. 000] 023026 RLCB -. . 0069883 -3.2949[. 002] 45516 LBACREIN(-1) . . 11174 4.0734[1000] 35439 LRUKYDS(-i) -. . 18796, -1.8854[. 067] 72199 LNIP(-1) -2.1545 . -2.9841[. 005] 0081279 ROV(-1) -. . 011105 -. 73189C. 4693 0041172 010118 RLCB(-1) . . . 40691[. 686] 10056 13451 LBACREIN(-2) . . . 74756[. 459] 20605 LRUKYDS(-2) . . 20082 1.0360[. 306] 1.2111 LNIP(-2) . 79996 1.5139[. 138] 0064698 010822 ROV(-2) . . . 59784[. 553] 0075966 010006 RLCB(-2) -. . -. 75923[. 452] 064061 12674 LBACREIN(-3) -. . -. 50561[. 616] 069630 19389 LRUKYDS(-3) -. . -. 35912[. 721] 66353 LNIP(-3) -. . 82167 -. 80754[. 424] 0033591 ROV(-3) -. . 010568 -. 31785[. 75] RLCB(-3) 0016658 010272 . . . 16218[. 872] LBACREIN(-4) 0068142 083401 081704[. 935] . . . 62356 30350 LRUKYDS(-4) . . 2.0546[. 046] 57672 53479 LNIP(-4) . . 1.0821[. 286] ROV(-4) 0027407 0079658 . . . 34406[. 733] 0027679 RLCB(-4) -. . 0076920 -. 35985[. 721] R-Squared . 99889 F-statistic F(24, 40) 1505.6[. 0003 R-Bar-Squared 99823 S. E. Regression . of . 025036 Squares Residual Sum of . 025076 Mean of Dependent Variable 10.2549 S. D. of Dependent Variable . 59526 Maximum of Log-lik elihood 163.2261 DW-statistic 1.8430 Diagno stic Tests * Test Statistics * LM Ve rsion * F Version * A: Serial Correlation *CHI-SO( 4)= 4.5860[. 3327*F( 4, 36)= 68319[. . 608]* * B: Functional Form *CHI-SO( 1)= 48629[. 486]*F( . 1, 39)= . 29397[. 591]* * C: Normality *CHI-SO( 2)= 6.4983[. 039]* Not applicable * D: Heteroscedasticity *CHI-SO( 1)= 1.8001[. 180]*F( 1, 63)= 1.7944[. 185]* A: Lagrange multiplier test of resid ual serial correlation RESET test the B: Ramsey's using squ are of the fitted val ues C: Based on a test of skewness and k urtosis of residuals the D: Based on regression of square d residuals on squared fitted values TABLE 10 Ordinary Least Squares Estimation Dependent variable is LBACREIN 55 observations used f or estimation f rom 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] CONST 1.3321 1.0402 1.2806[. 210] 34137 LRUKYDS -. . 27006 -1.2640[. 216] LNIP 2.1198 . 50171 4.2251[. 000] 030662 ROV . . 0086020 3.5646[. 001] 021049 RLCB -. . 0079781 -2.6383["013] 42233 LBACREIN(-1) . . 11089 3.8085[. 001] 53869 LFUKYDS(-1) -. "22082" -2.4395[. 021] 75597 LNIP(-1) -2.3248 . -3. "0753[. 004] 013123 76643[. 449] ROY(-1) -. 010058 . -. 0048474 40000[. 692] RLCB(-1) . . 012119 . 023757 862] LBACREIN(-2) . . 13578 . 17496[. 33485 21668 133] LRUK: YDS(-2) . . 1.5454[. LNIP(-2) 1.2057 . 84065 1.4343[. 162] 026803 ROV(-2) . . 014883 1.8010[. 082] 024553 086] PLCB(-2) -. . 013842 -1.7738[. 018112 12954 13983[. 890] LPACFEIN(-3) -. . -. 28696 202] LRUKYDS(-3) -. . 22011 -1.3037[. 82145 057] LNIP(-3) -1.6247 . -1.9779[. 013385 014312 357] ROV(-3) -. . -. 93527[. 0099275 013445 73836[. 466] FLCB(-3) . . . 073450 083853 388] LBACFEIN(-4) . . . 87595[. 93541 29026 003] LRUKYDS(-4) . . 3.2227[. 1.2769 53668 2.3792[. 024] LNIP(-4) . 012345 013990 385] ROV(-4) . . . 88245[. FLCD(-4) -. 011439 . 012860 -. 88946[. 381] R-Squared . 99881 F-statistic F(24, 30) 1045.8[. 000] 99785 S. E. 022409 P-Bar-Squared . of Regression . Residual Sum of Squares . 015066 Mean of Dependent Variable 10.0879 S. D. of Dependent Variable . 48341 Maximum of Log-lik elihood 147.5321 DW-statistic 2.1383 Diagnostic Tests * Test Statistics * LM Version * F Version Correlation *CHI-SO( 4)= 4.1881[. 381]*F( * A: Serial 4, 26)= . 53576C. 711]X Form SCHI-SO( 1)= 54545[. 460]*F( 29048[. * B: Functional . 1, 29)= . 594]) * C: Normality *CHI-SO( 2)a 2.4638[. 292] Not applicable X *CHI-SO( 1)= 13703[. 711]*F( * D: Heteroscedasticity . 1, 53)= . 13239[. 717J * E: Predictive Failure *CHI-SO( 10)= 19.9344[. 030]*F( 10, 30)= 1.9934[. 071]ä A: Lagrange multiplier test of residual serial correlatio n D: Ramsey's RESET test using the square of the fitted val ues C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on square d fitted values E: A test of adequacy of predictions (Chow's second test) TABLE 11 Variable Deletion Test (OLS case) Dependent variable is LBACREIN List of the variables deleted from the regressions LRUKYDS(-1) LNIP(-1) ROV(-1) RLCP(-1) LBACREIN(-2) LRUKYDS(-2) LNIP(-2) ROV(-2) RLCB(-2) LBACREIN(-3) LRUKYDS(-3) LNIP(-3) ROV(-3) RLCB(-3) LBACREIN(-4) LRUKYDS(-4) LNIP(-4) ROV(-4) RLCB(-4) 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] 80853 CONST . . 52561" 1.5383[. 130] 23453 LRUKYDS . . 092817 2.5268[. 015] 69678 LNIP . . 072310 9.6360[. 000] 028274 ROV . . 0063952 4.4212[. 000] 024785 RLCB -. _ . 0055098 -4.4983[. 000] 45151 LBACREIN(-1) . . 054591 8.2708[. 000] Joint test of zero restrictions on the coefficient of deleted variables: Lagrange Multiplier Statistic CHI-SQ(19)= 33.3536[. 022] Likelihood Ratio Statistic CHI-SQ(19)= 51.2871[. 000] F Statistic F(19,30)= 2.4329[. 014] Y TABLE 12 Variable Deletion Test (OLS case) Dependent variable is LBACREIN List of the variables deleted from the regression: LFUKYDS(-1) LNIP(-1) ROV(-1) RLCB(-1) LBACREIN(-2) LRUKYDS(-2) LNIP(-2) ROV(-2) RLCB(-2) LBACREIN(-3) LRUKYDS(-3) LNIP(-3) ROV(-3) PLCB(-3) LBACREIN(-4) LRUKYDS(-4) LNIP(-4) ROV(-4) RLCE(-4) 65 observations used for estimation from 7502 to 9102 Regressor Coefficient Standard Error T-Fatio[Prob] 79068 CONST . . 50182 1.5756[. 120] 19881 LRUKYDS . . 082961 2.3964[. 020] 64950 LNIP . . 067195 9.6660[. 000] 032855 7.5303[. ROV . . 0043631 000] 028242 RLCB -. . 0040932 -6.8997C. 000] LBACREIN(-1) . 49464 . 048295 10.2420[. 000] Joint test of zero restrictions on the coefficient of deleted variables: Lagrange Multiplier Statistic CHI-SO(19)= 29.6848[. 056] Likelihood Ratio Statistic CHI-SQ(19)= 39.6548C. 004] F Statistic F(19,40)= 1.7696[. 064] TADLE 13 Ordinary Least Squares Estimation Dependent variable is LBACREIN 65 observations used for estimation from 7502 to 9102 Regressor Coefficient Standard Error T-Ratio[Prob] 79068 50162 CONST . . 1.5756[. 120] 19881 082961 LRUKYDS . . 2.3964[. 020] 64950 LNIP . . 067195 9.6660[. 000] 032855 ROV . . 0043631 7.5303[. 000] 028242 RLCB -. . 0040932 -6.8997[. 000] 49464 LE R-Squared . 99796 F-statistic F( 5,59) 5786.0[. 000] 99779 S. E. Regression 027969 R-Bar-Squared . of . Residual Sum of Squares . 046155 Mean of Dependent Variable 10.2549 Variable 59526_ S. D. of Dependent . Maximum of Log-likelihood 143.3988 DW-statistic 1.7666 Durbin's h-statistic 1.0213[. 307] Diagnostic Tests * Test Statistics # LM Version *F Version * * *' * A: Serial Correlation *CHI-SQ( 4)= 4.4276[. 3513*F( 4,55)= 1.0051[. 413]4 * * 4 * B: Functional Form *CHI-SQ( 1)= 2.3694[. 124]#F( 1,58)= 2.1942[. 144]* * * # 2)= * C: Normality *CHI-SQ( . 34882[. 640]* Not applicable * * * k D: Heteroscedasticity *CHI-SQ( 1)= 047012[. 828]*F( 045599[. * . 1,63)= . E. 214 A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 14 Ordinary Least Squares Estimation Dependent variable is LBACREIN 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Frob] 80853 CONST . . 52561 1.5383[. 130] 23453 LRUKYDS . . 092817 2.5268[. 015] 69678 LNIP . . 072310 9.6360[. 000] 028274 ROV . . 0063952 4.4212[. 000] 024765 RLCB -. . 0055098 -4.4983[. 000] 45151 LDACREIN(-1) . . 054591 8.2708[. 000] R-Squared . 99697 F-statistic F( 5,49) 3220.9[. 000] 99666 S. E. Regression R-Bar-Squared . of . 027950 Residual Sum of Squares . 038279_ Mean of Dependent Variable 10.0879 S. D. of Dependent Variable . 48341 Maximum of Log-likelihood 121.8885 DW-statistic 1.7188 Durbin's h-statistic 1.1405[. 254] Diagnostic Tests * Test Statistics # LM Version *F Version Correlation *CHI-SO( 4)= 4.2798[. 369]*F( * A: Serial 4,45)= . 94928[. 4457' * B: Functional Form *CHI-SO( 1)= 1.4592[. 227]*F( 1,48)= 1.7,082C. 2581' *CHI-SQ( 2)= * C: Normality . 26491[. 676]* Not applicable Heteroscedasticity *CHI-SO( 1)= 018527[. 892]#F( 017859[. 894]) K D: . 1,53)= . * E: Predictive Failure SCHI-SQ( 10)= 10.0813[. 4333*F( 10,49)= 1.0081["450]) **#> Test SCHI-SO( 6)= 5.1444[. 525]*F( 85740[. * F: Chow 6,53)= . 532]> A: Lagrange multiplier test of residual serial correlation E; Famsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) F: Test of stability of the regression coefficients TABLE 15 Ordinary Least Squares Estimation Dependent variable is LBACREIN 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] 28029 CONST . . 76062 . 37902[. 706] 43894 056237 LDACREIN(-1) . . 7.8052[. 000] 025950 ROV . . 0068564 3.7848[. 000] 022293 RLCB -. . 0061109 -3.6480[. 001] 68937 LNIP . . 072807 9.4685[. 000] 33670 LRUKYDS . . 14237 2.3650[. 022] 19467 20553 94716[. 348] UHAUS -. . -. R-Squared . 99702 F-statistic F( 6,48) 2678.6[. 000] 99665 S. E. Regression 027979 R-Bar-Squared . of . Residual Sum of Squares . 037577 Mean of Dependent Variable 10.0879 S. D. of Dependent Variable . 48341 Maximum of Log-likelihood 122.3977 DW-statistic 1.6476 Durbin's h-statistic 1.4376[. 151] Diagnostic Tests * Test Statistics * LM Version *F Version Correlation *CHI-SQ( 4)= 4.3038[. 3663*F( * A: Serial 4,44)= . 93383[. 4533) **#a Functional Form *CHI-SO( 1)= 56896[. 451]*F( 4871) * B: . 1,47)= . 49129[. ***y 2)= C: Normality *CHI-SQ( . 42192[. 810]* Not applicable Heteroscedasticity SCHI-SO( i)= 070770[. 790]*F( * D: . 1,53)= . 068285[. 795]r A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Dased on the regression of squared residuals on squared fitted values TABLE 16 Ordinary Least Squares Estimation Dependent variable is LBACROIN 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] 31460 99482 31643[. 753] CONST . . . 98688 LBACROIN(-1) . . 031195 31.6360[. 000] 0074258 ROV -. . 0060913 -1.2191[. 229] 0088873 TREBIRA . . 0061193 1.4523[. 153] 084783 LNIP . . 034414 2.4637[. 017] 073717 LRUKYDS -. . 18606 -. 39621[. 694] 15089 23607 63915[. UHAUS . . . 526] R-Squared . 99921 F-statistic F( 6,48) 10158.5[. 000] 99911. S. E. Regression R-Dar-Squared . of . 023968 Residual Sum of Squares . 027575 Mean of Dependent Variable 10.9378 S. D. of Dependent Variable . 80556 Maximum of Log-likelihood 130.9086 DW-statistic 1.5652 Durbin's h-statistic 1.6571[. 097] Diagnostic Tests * Test Statistics * LM Version *F Version A: Serial Correlation *CHI-SQ( 4)= 2.8406[. 5857*F( * 4,44)= . 59905[. 665]1 * # Form *CHI-SO( i)= 22032C. 639]*F( * B: Functional . 1,47)= . 18903[. 6663) * C: Normality *CHI-SO( 2)= 4.7468C. 093]ß Not 'applicable * * * > *CHI-SQ( i)= 22464[. 636]*F( * D: Heteroscedasticity . 1,53)= . 21736[. 643]) A: Lagrange multiplier test of residual serial correlation E: Samsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 17 Ordinary Least Squares Estimation Dependent variable is LBACRTO 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Fatio[Prob] CONST -2.4260 . 97715 -2.4828[. 017] 87322 LDACRTO(-1) . . 040281 21.6785[. 000] 0058994 ROV -. . 0065055 -. 90684[. 369] 0056996 0065780 391] TREBIRA . . . 86646[. 16102 LNIP . . 042480 3.7904[. 000] 47695 LRUKYDS . . 19317 2.4691[. 017] 40993 UHAUS -. . 24567 -1.6686[. 102] R-Squared . 99883 F-statistic F( 6,48) 6845.6[. 000] 99869_ S. E. R-Dar-Squared . of Regression . 025777 Residual Sum of Squares . 031894 Mean of Dependent Variable 11.3063 71132 S. D. of Dependent Variable . Maximum of Log-likelihood 126.9072 DW-statistic 1.5823 Durbin's h-statistic 1.6232[. 105] Diagnostic Tests * Test Statistics * LM Version ýK F Version Correlation *CHI-SO( 4)= 1.2853[. 864]*F( * A: Serial 4,44)= . 2632[. 900]) **#> Form *CHI-S(3( 1)= 69466[. 405]#F( 60121[. 4427) * P: Functional . 1,47)= . 2)= 020663[. * C: Normality *CHI-SQ( . 990]* Not applicable * D: Heteroscedasticity *CHI-SQ( 1)= 1.0975[. 295]*F( 1,53)= 1.0791[. 304]) A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Erased on a test of skewness and kurtosis of residuals D: E{ased on the regression of squared residuals on squared-fitted values TABLE 18 Ordinary Least Squares Estimation Dependent variable is LBACREMA 55 observations used for estimation from 7601 to 8903 Regressor Coefficient Standard Error T-Ratio[Prob] 35933 CONST -. . 74550 -. 48200[. 633] LIPGNP -. 16194 1.0891 -. 14869["883] 017869 016456 RCRCA . . 1.0858[. 286] 0022717 GTRBR -. . 017372 -. 13077[. 897] 035235 43417 LRGGNP -. . -. 081156[. 936] 57581 21446 LDACREMA(-1) . . 2.6849[. 012] 063215 78316 LIPGNP(-1) -. . -. 080717[. 936] 022368 RCRCA(-i) -. . 020045 -1.1159[. 273] GTRBR(-1) 0053293 019135 . . . 27851[. 783] LR6GNP(-1) 0065390 38454 . . . 017005[. 987] LBACREMA(-2) 030388 24024 . . . 12649[. 900] LIPGNP(-2) 16640 69188 . . . 24051[. 812] 024663 RCRCA(-2) . . 019380 1.2726[. 213] 033598 GTRBR(-2) -. . 020907 -1.6070[. 119] 0011550 37284 LRGGNP(-2) -. . -. 0030979[. 998] 31913 25230 LBACREMA(-3) -. . -1.2649[. 216] 65022 70296 LIPGNP(-3) -. . -. 92498[. 362] 010640 019360 RCRCA(-33) -. . -. 54960[. 587] 039694 GTRBR(-3) . . 020788 1.9095[. 066] 46122 LRGGNP(-3) . . 39312 1.1732[. 250] LBACREMA(-4) 18037 24441 73797[. . . . 466] 1.1508 LIPGNP(-4) . 78096 1.4736[. 151] RCRCA(-4) 0050343 011643 . . . 50111[. 620] 016251 016112 GTRBR(-4) -. . -1.0086[. 321] LRGGNP(-4) 25019 46309 54027[. . . . 593] R-Squared . 99248 F-statistic F(24, 30) 164.9864[. 000] R-Dar-Squared . 98647 S. E. of Regression 026151 Squares . Residual Sum of . 020516 Mean of Dependent Variable 5.1154 S. D. of Dependent Variable . 22478 Maximum of Log-lik elihood 139.0401 DW-statistic 2.0718 Diagnostic Tests * Test Statistics * LM Version * F Version * A: Serial Correlation *CHI-SQ( 4)= 6.9427[. 1393*F( 4, 26)= . 93904[. 457]4 * P: Functional Form *CHI-SO( 1)= 11.00801.001]*F( 1, 29)= 7.2566[. 012]1 *CHI-SO( * C: Normality 2)= 67.4654[. 000] Not applicable x * D: Heteroscedasticity *CHI-SO( 1)= 096699[. 756]*F( 1, 53)= . . 093347[. 761]1 * E: Predictive Failure *CHI-SQ( 5)= 1.7945[. 877]*F( 5, 30)= . 35889[. 872]4 test A: Lagrange multiplier of residual serial correlatio n RESET test the D: Ramsey's using square of the fitted val ues C: Based on a test of skewness and kurtosis of residuals the D: Based on regression of squared residuals on square d fitted values E: A test of adequacy of predictions (Chow's second test) TABLE 19 Ordinary Least Squares Estimation Dependent variable is LBACREMA 55 observations used for estimation from 7601 to 8903 Regressor Coefficient Standard Error T-Ratio[Prob] 23809 75921 756] CONST -. . -. 31361[. LIPGNP -1.2089 1.2004 -1.0070[. 323] 023049 177] RCRCA . . 016634 1.3856[. 0051042 GTRBR -. . 017002 -. 30022[. 766] 48238 323] LRGGNP -. . 47900 -1.0071[. 61491 LBACREMA(-1) . . 22899 2.6854[. 012] 1.0086 417] LIPGNP(-i) 1.2226' . 82490[. 035924 RCRCA(-1) -. . 020629 -1.7414[. 093] 0075801 690] GTRBR(-1) . . 018805 . 40308[. 62724 LRGGNP(-1) . . 55312 1.1340[. 267] 12237 641] LFACREMA(-2) . - . 25957 . 47145[. 52754 651] LIPGNP(-2) . 1.1527 . 45764[. 029605 RCRCA(-2) . . 020181 1.4669[. 154] 029204 167] GTRBR(-2) -. . 020540 -1.4218[. 064044 902] LRGGNP(-2) . . 51602 . 12411C. 24648 28234 LBACREMA(-3) -. . -. 87300[. 390] LIPGNP(-3) -. 24266 1.1130 -. 21620[. 829] 0086810 669] RCRCA(-3) -. . 020072 -. 43248[. 034192 GTRBR(-3) . . 020487 1.6690[. 107] 77873 56441 179] LRGGNP(-3) . . 1.3797[. 014206 LBACREMA(-4) -. . 25770 -. 055126[. 956] 33456 LIPGNP(-4) . . 92838 . 36038[. 721] 0088733 011758 457] RCRCA(-4) . . . 75468[. 016762 015929 302] GTRBR(-4) -. . -1.0523[. 28630 LRGGNP(-4) -. . 52506 -. 54527[. 590] 20508 S1 -. . 10690 -1.9183[. 066] 11209 S2 -. . 078295 -1.4317[. 164] 12051 S3 -. . 103,52 -1.1641[. 255] R-Squared . 99356 F-statistic F(27, 27) 154.2491[. 000] 98712 S. E. Regression R-Bar-Squared . of . 025513 Residual Sum of Squares . 017574 Mean of Dependent Variable 5.1154 Variable 22478 S. D. of Dependent . Maximum of Log-lik elihood 143.2961 DW-statistic 2.1119 Diagnostic Tests * Test Statistics # LM Version # F Version * A: Serial Correlation *CHI-SO( 4)= 14.7266[. 0053*F( 4, 23)= 2.1026[. 113]* * B: Functional Form *CHI-SO( 1)= 12.7089[. 000]*F( 1, 26)= 7.8133[. 010] * C: Normality *CHI-SO( 2)= 56.9649[. 000]* Not applicable Heteroscedasticity *CHI-SLR( 1)= 24830[. 616]*F( * D: . 1, 53)= . 24036[. 626] Predictive Failure *CHI-SLR( 5)= 4.1430[. 529]*F( * E: 5, 27)= . 82860[. 541]A A: Lagrange multiplie r test of residual serial correlation B: Ramsey's RESET tes t using the square of the fitted val ues C: Based on a test of skewness and kurtosis of residuals D: based on the regre ssion of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) TABLE 20 Ordinary Least Squares Estimation Dependent variable is LBACREMA 36 observations used for estimation from 8004 to 8907, Regressor Coefficient Standard Error T-Ratio[Prob] 1.7754 CONST . 88080 2.0157[. 053] 024272 0073592 RCRCA -. . -3.2981[. 003] 031603 GTRBR . . 0081176 3.0932C. 0013 17755 LIPGNP -. . 30159 -. 58871[. 560] 61163 LRGGNP . . 16511 3.7045[. 001] 66422 091883 LBACREMA(-1) . . 7.2289[. 000] 96048 R-Squared . F-statistic F( 5,30) 145.81200.000] 95389 S. E. Regression 019172 R-Bar-Squared . of . 011027 Residual Sum of Squares . Mean of Dependent Variable 5.2647 S. D. of Dependent Variable . 089284 - Maximum of Log-likelihood 94.5550 DW-statistic 1.6201 Durbin's h-statistic 1.3660[. 17] Diagnostic Tests * Test Statistics * LM Version #F Version * * * ' * A: Serial Correlation *CHI-SO( 4)= 6.51-30[. 164]*F( 4, 26)= 1.4357[. 250J> * * * > * B: Functional Form *CHI-SO( 1)= 4.6352[. 031]*F( 1, 29)= 4.2857[. 0473) 2)= 93140[. ýc C: Normality *CHI-SQ( . 628] Not applicable * * * ' * D: Heteroscedasticity SCHI-SO( 1)= 1.0544[. 305]*F( 1, 34)= i. O258[. 318]ý * * * > Predictive Failure *CHI-SQ( 5)= 2.4704[. 781]*F( 49408[. 778] * E: 5, 30)= . A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) .ý TABLE 21 Ordinary Least Squares Estimation Dependent variable is LBACREMA 22 observations used for estimation from 7502 to 8003 Regressor Coefficient Standard Error T-Ratio[Prob] 62832[. 539] CONST . 52153 . 83003 . 014915 0089298 1.6703[. 114] RCRCA . . 0075008. 59227[. 562] GTRBR -. . 012664 -. 32936 137] LIFGNP . 51499 . 1.5636[. 051] LRGGNP . 47562 . 22586 2.1058[. 26919 15098 1.7829[. 094] -LSACFEMA(-1) . . 000] R-Squared . 97435 F-statistic F( 5,16) 121.5760[. 96634 S. E. R-Ear-Squared . of Regression . 015561 Sum Squares 0038742 Mean of Dependent Variable 4.8197 Residual of . - S. D. of Dependent Variable . 084815 Maximum of Log-likelihood 63.8724 1.7404 Durbin's 389] DW-statistic h-statistic . 86227[. Diagnostic Tests * Test Statistics * LM Version *F Version * A: Serial Correlation *CHI-SQ( 4)= 6.1201[. 190]*F( 4,12)= 1.1562C. 3787* * D: Functional Form *CHI-SO( 1)= 1.5005[. 221]*F( 1,15)= 1.0980[. 311]* **** * C: Normality *CHI-SO( 2)= 1.1755[. 556]* Not applicable **** 1)= 83048[. * D: Heteroscedasticity *CHI-SQ( . 36]*F( 1,20)= . 78459E. 396]* **** * E: Predictive Failure *CHI-SQ( 41)= 133.3209[. 000]*F( 41,16)= :. 2517[. 007]4 ***A * FsChow Test *CHI-SO( 6)= 65.6391["000]*F( 6,51)= 10.9399[. 000]* A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on"a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) F: Test of stability of the regression coefficients TABLE 22 Ordinary Least Squares Estimation Dependent variable is LRGGNP 36 observations used for estimation from 8004 to 8903 Regressor Coefficient Standard Error T-Ratio[Prob] 343] CONST 1.3197 1.3687 . 96417[. 027380 35973 940] LFGGNP(-1) . . . 076114[. 62398 47679 201] LIPGNP(-1) -. . -1.3087[. GTFDR(-1) -. 0076095 . 010058 -. 75658[. 455] 35045 060J LNMi(-1) . . 17960 1.9513[. 20111 303J LDACREMA(-1) . . 19184 1.0483[. R-Squared . 69364 F-statistic F( 5,30) 13.5849[. 000] 64258 R-Bar-Squared . S. E. of Regression . 036468 Residual Sum of Squares . 039898 Mean of Dependent Variable 1.5234 060999 " S. D. of Dependent Variable . Maximum of Log-likelihood 71.4075 DW-statistic 2.5186 Durbin's h-statistic *NONE* Diagnostic Tests * Test Statistics * LM Version *F Version * A: Serial Correlation *CHI-SO( 4)= 32.3701[. 000]#F( 4,26)= 57.9650[. 000]3 ***M * B: Functional Form *CHI-SQ( 1)= 5.2221[. 022]*F( 1,29)= 4.92^05[. 035]4 * C: Normality *CHI-SO( 2)= 1.8534[. 396] Not applicable 9 * D: Heteroscedasticity *CHI-SQ( i)= 1.3857[. 239]*F( 1,34)= 1.3611[. 25174 *** ýI Failure *CHI-SO( 5)= 3.3951[. 639]*F( 5,3O)= 67902[. 643]1 * E: Fredictive . A: Lagrange multiplier test of residual serial correlation D: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) TABLE 23 Ordinary Least Squares Estimation Dependent variable is LBACREMA 36 observations used for estimation from 8004 to 8903 Regressor Coefficient Standard Error T-Ratio[Frob] CONST . 31470 . 81014 . 38844E. 701] 97201 16296 LRGGNF . . 5.9647[. 000] 011416 RCRCA -. . 0068451 -1.6678[. 106] 029543 GTRBR . . 0066695 4.4296[. 000] 49955 LIFGNR . . 30035 1.6632[. 107] 22162 LBACREMA(-1) . . 13466 1.6458[. 111] 81282 HREBCMA -. . 20506 -3. '9639[. 000] R-Squared . 97437 F-statistic F( 6,29) 183.7181[. 000] 96906 R-Bar-Squared . S. E. of Regression 015704 squares , . Residual sum of . 0071520 Mean of Dependent Variable 5.2647 6f S. D. of Dependent Variable . 089284 Maximum Log-likelihood 102.3481 DW-statistic 1.7104 Durbin's h-statistic 1.4745[. 140] Diagnostic Tests * Test Statistics * LM Version *F Version Correlation *CHI-SQ( 4)= 3.8715[. 424]*F( 4,25)= 75312[. 565]4 * A: Serial . * * 8 Form *CHI-SQ( 1)= 15274[. 696]*F( 732]4 * B: Functional . 1,28)= . 11931[. * * * A * C: Normality *CHI-SO( 2)= 1.2672E. 531]* Not applicable * * * A Ný D: Heteroscedasticity *CHI-SO( 1)= 1.5417[. 214]*F( 1,34)= 1.5211[. 226]I A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: E(ased on a test of skewness and kurtosis of residuals' D: Based on the regression of squared residuals on squared fitted values TABLE 24 Ordinary Least Squares Estimation Dependent variable is LBACREMA 36 observations used for estimation from 8004 to 8903 Regressor Coefficient Standard Error T-Fatio[Prob] 60632 86523 70076[. 490] CONST . . . 1.0240 15837 6.4658[. LRGGNP . 000] 013911 0092921 146] RCRCA -. . -1.4971[. 031021 GTRBF . . 0086344 3.5927[. 001] 36052 LIPGNP . . 34543 1.0437[. 306] 27797 LBACFEMA(-1) . . 16886 1.6462[. 112] 013442 026918 49937[. 622] Si -. . -. S2 -. 0022458 . 020278 -. 11075[. 913] 017970 S3 -. . 018340 -. 97982[. 336] HFEBCMA 91843 28692 004] -. _ . -3.2010[. R-Squared . 97893 F-statistic F( 9,26) 134.2007[. 000] 97163 S. E. Regression R-Bar-Squared . of . 015030 0058795 Mean Residual Sum of Squares . of Dependent Variable 5.2647 089284 S. D. of Dependent Variable . Maximum of Log-likelihood 105.8748 DW-statistic 1.5912 Durbin's h-statistic *NONE* Diagnostic Tests * Test Statistics * LM Version #F Version ************#******##************#********************************************> ***> *CHI-SQ( 4)= 2 . 314[. 693]*F( 36343[. 832]) * A: Serial Correlation 4,22)= . *CHI-SO( 1)= 034942[. * B: Functional Form . 852]*F( 1,25)= . 024289[. 8771) * C: Normality *CHI-SQ( 2)= 1.6854[. 431] Not applicable **#> *CHI-SQ( 1)= 040879[. 840]*F( * D: Heteroscedasticity . 1,34)= . 038652[. 845]) A: Lagrange multiplier test of residual serial correlation D: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 25 Instrumental Variable Estimation Dependent variable is LE'ACREMA List of instruments: CONST LRGGNP(-1) RCRCA(-1) GTRBR(-1) LIPGNP(-i) LBACREMA(-2) 36 observations used for estimation from 8004 to 8903 Regressor Coefficient Standard Error T-Ratio[Prob] CONST 6.4250 5.9995 1.0709[. 293] 1.3245 LRGGNP . 77588 1.7071[. 098] 040922 187] RCRCA -. . 030294 -1.3508[. 025565 GTRBR . . 020619 1.2399[. 225] LIPGNP -2.0921 2.4379 -. 85817[. 398] 1.2853 LBACREMA(-1) . 64504 1.5210[. 139] R-Squared . 88905 F-statistic F( 5,30) 48.0766[. 000] 87055 S. E. Regression R-Bar-Squared . of . 03212~ Residual Sum of Squares . 030957 Mean of Dependent Variable 5.2647 Variable 069284 Value S. D. of Dependent . of IV Minimand . 0000 DW-statistic 2.3399 Sargan's *NONE* Diagnostic Tests * Test Statistics * LM Version *F Version * * # r * A: Serial Correlation *CHI-SO( 4)= 1.3477C. 853]ß Not applicable * * * r * D: Functional Form *CHI-SQ( 1)= i. 1632[. 281]* Not applicable * C: Normality *CHI-SO( 2)= 4. ý598[. 113]* Not applicable * D: Heteroscedasticity *CHI-SQ( 1)= 1. Q959[. 295]ß Not applicable A: Lagrange multiplier test of residual serial correlation D: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 26 Instrumental Variable Estimation Dependent variable is LBACREMA List of instruments: CONST LRGGNP(-1) RCRCA(-1) GTRBR(-1) LIRGNF(-1) LBACREMA(-2) S1 S2 S3 36 observations used for estimation from 8004 to 8903 ****************#****#*************#* *********#**********************#*#******> Regressor Coefficient Standard Error T-Ratio[Prob] CONST 4.7372 4.6934 1.0093[. 322] 1.0147 LRGGNF . 30879 3.2862[. 003] 038867 RCRCA -. . 036253 -1.07210.2933 032137 GTRBR . . 016613 1.9345[. 064] LIPGNP -1.4645 2.1479 -. 68181[. 501] LBACREMA(-1) 1.1416 1.0558 1.0813C. 289] 022185 Si -. . 10087 -. 21993[. 828] 8157 S2 -. 018889 . 079875 -. 21648[. S3 -. 037288 . 075763 -. 49216[. 627] 94522 R-Squared . F-statistic F( 8r 27) 58.2307[. 000] 92898 S. E. Regression 023793 R-Bar-Squared . of . Residual Sum of Squares . 015285 Mean of Dependent Variable 5.2647 Variable 089284 Value IV Minimand 0000 S. D. of Dependent . of . DW-statistic 2.3421 Sargan's *NONE* Diagnostic Tests * Test Statistics * LM Version *F Version * * * * A: Serial Correlation *CHI-SO( 4)= 3.9102E. 418]* Not applicable * * * * B: Functional Form *CHI-SO( 1)= 1.5723[. 210]* Not applicable * * * * C: Normality *CHI-SO( 2)= 3.2706[. 195]* Not applicable * * * * D: Heteroscedasticity *CHI-SO( 1)= 2.4764[. 116] Not applicable A: Lagrange multiplier test of residual serial correlation D: Ramsey's RESET test using the square of the fitted values C; Eased on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 27 Instrumental Variable Estimation Dependent variable is LBACREMA List of instruments: CONST LRGGNP(-1) RCRCA(-1) - RTDEP(-1) LIPGNP(-1) LBACREMA(-2) S1 S2 S3 36 observations used for estimation from 8004 to 6903 Regressor Coefficient Standard Error T-Ratio[Prob] CONST -4.5984 6.2938 -. 73062[. 471] 1.0409 LRGGNP . 93171 1.1172[. 274] 037556 RCRCA . . 030528 1.2302[. 229] 0045315 033732 13434[. 894] RTDEP . . . LIPGNP 2.9247 2.7143 1.0775[. 291] L8ACREMA(-1) -1.0837 1.0883 -. 99574[. 328] 18857 S1 . . 095787 1.9686[. 059] 14952 - S2 . . 075199 1.9883[. 057] 12135 S3 . . 074643 1.6257[. 116] R-Squared . 90377 F-statistic F( 8,27) 31.6970[. 000] 87526 S. E. Regression 031534 R-Bar-Squared . of . Residual Sum of Squares . 026849 Mean of Dependent Variable 5.2647 Dependent Variable 089284 Value IV Minimand S. D. of . of . 0000 DW-statistic 1.3765 Sargan's *NONE* Diagnostic Tests * Test Statistics * LM Version *F Version * * * a * A: Serial Correlation *CHI-SQ( 4)= 2.8407[. 585] Not applicable * B: Functional Form *CHI-SQ( 1)= , 2.6979[. 100]* Not applicable * * * 8 * C: Normality *CHI-SO( 2)= 2.7898[. 248]* Not applicable * D: Heteroscedasticity *CHI-SO( 1)= 5.4182[. 020]* Not applicable A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 28 Ordinary Least Squares Estimation Dependent variable is LBACREMA 36 observations used for estimation from 80Q4 to 8903 Regressor Coefficient Standard Error T-Ratio[Prob] -CONST -. 19398 1.0125 -. 19158[. 850] LRGGNP . 77563 . 23901 3.2452[. 003] 6703E-3 010741 062403[. 951] RCRCA -. . -. 013468 RTDEP . . 0099354 1.4556[. 186] LIPGNP . 79938 . 38656 2.0679[. 048] 095496 16327. LBACREMA(-1) . . . 56490[. 5637 064207 020762 S1 . . 3.0925[. 005] 055613 016281 62 . . 3.4281[. 002] S3 . 034212 . 015014 2.2786[. 031] ***************###*********##*#**#*#****#*****#*******************************> R-Squared . 96742 F-statistic F( 8,27) 100.2310[. 000] 95777 S. E. Regression R-Bar-Squared . of . 018347 0090887 Residual Sum of Squares . Mean of Dependent Variable 5.2647 089284 S. D. of Dependent Variable . Maximum of Log-likelihood 98.0347 DW-statistic 1.2204 Durbin's h-statistic 11.6426[. 000] Diagnostic Tests * Test Statistics # LM Version *F Version * A: Serial Correlation SCHI-SQ( 4)= 7.4930[. 1123*F( 4, 23)= 1.5114[. 232]a * * .* K B: Functional Form SCHI-SO( i)= 3. O715[. OBO]*F( 1, 26)= 2.4252[. 1.3113) * * * * C: Normality *CHI-SQ( 2)= . 23515[. 889] Not applicable I * D: Heteroscedasticity *CHI-SQ( 1)= 2.1698[. 141]*F( i, 34)= 2.1BO7[. 149]1 Failure *CHI-SO( 5)= 1881311.00]*F( * E: Fredictive . 5, 27)= . 037626[1. GG]x A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Dased on a test of skewness and kurtosis of residuals D; Based on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) TABLE 29 Ordinary Least Squares Estimation Dependent variable is LNM1 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Fatio[Prob] CONST 2.3506 2.1862 1.0752[. 90] 30430 37295 81594[. 421] LRUKYDS . . . 0027194 TREBIRA -. . 0030928 -. 87927[. 386] 13397 LNIP . . 52997 . 25278[. 802] 017464 034059 612] S1 . . . 51275[. 82 . 061714 . 038245 1.6136[. 116] 027178 33 . . 028624 . 94949[. 349] 80488 16954 LNMI(-1) . . 4.7475[. 000] 035825 35655 LRUKYDS(-1) -. . -. 10048[. 921] TPEBIRA(-1) -. 0019011 . 0041827 -. 45450[. 653] 10027 80507 902] LNIP(-1) . . . 12455[. 10112 22356 45234[. 654] LNM1(-2) . . . 56339 LRUKYDS(-2) -. . 33305 -1.6916[. 1C'OJ 3665E-3 0039493 TREBIRA(-2) . . . 092806[. 927] 62832 76441 82197[. LNIP(-2) -. . -. 417] 096791 23172 41771[. 679] LNM1(-3) -. . -. 16949 LFUKYDS(-3) -. . 33620 -. 50415[. 618] 9524E-3 0038873 8083 TREBIRA(-3) -. . -. 24499[. 72150 78914 91428[. 367] LNIP(-3) . . . 32485 18865 LNM1(-4) . . 1.7219[. 095] 052257 LRUKYDS(-4) -. . 32779 -. 15941[. 874] 0033477 TREDIFA(-4) . . 0031951 1.0478[. 303] 38369 LNIP(-4) -. . 47425 -. 80904[. 424] R-Squared . 99885 F-statistic F(22, 32) 1267.7[. 000] 99807 S. E. Regression R-Bar-Squared . of . 024629 Residual Sum of Squares . 019411 Mean of Dependent Variable 10.5365 S. D. of Dependent Variable . 56005 Maximum of Log-lik elihood 140.5631 DW-statistic 1.8263 Diagnostic Tests * Test Statistics * LM Version ýk F Version Correlation SCHI-SO( 4)= 1.9701[. 741]*F( * A: Serial 4, 28)= . 26005[. 901]1 Functional Form *CHI-SD( 1)= 094030[. 759]*F( * B: . 1, 31)= . [)53090[. 819]1 * C: Normality *CHI-SO( 2)= 1.6194[. 445] Not applicable * D: Heteroscedasticity *CHI-SO( 1)= 1.3574[. 44]*F( 1, 53)= 1.3411[. 25 ]8 ***#*ýk*********#****** **#*#************************#************************** A: Lagrange multiplie r test of residual serial correlation B: Ramsey's RESET tes t using the square of the fitted val ues C: Eiased on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 30 Variable Deletion Test (OLS case) Dependent variable is LNM1 List of the variables deleted from the regression: LNM1(-2) LNM1(-3) LNMI(-4) TREBIRA(-1) TREBIRA(-2) TREBIRA(-3) TREBIRA(-4) LNIF(-1) LNIP(-2) LNIP(-3) LNIP(-4) LRUKYDS(-1) LRUKYDS(-2) LRUKYDS(-3) LRUKYDS(-4) 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Frob] CONST -1.2430 1.0899 -1.1405[. 260] 93225 LNM1(-1) . . 057709 16.1543[. 000] 0041181 0016314 415] "TREBIRA -. . -2.5243[. 034486 LNIP . . 040625 . 84888[. 4003 28016 229] LRUKYDS . . 22985 1.2189[. 019253 S1 -. . 016447 -1.1706[. 248] 020343 S2 . . 015896 1.2798[. 247] 5155E-4 88.3 -. . 011712 -. 0044010[. 997] Joint test of zero restrictions on the coefficient of deleted variables: Lagrange Multiplier Statistic CHI-SQ(15)= 17.2703[. 303] Likelihood Ratio Statistic CHI-SQ(15)= 20.7287[. 146] F(15,32)= 97651[. F Statistic . 500] TABLE 31 Ordinary Least Squares Estimation Dependent variable is LNMI 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] CONST -1.2430 1.0899 -1.1405[. 260] 93225 057709 16.1543[. 000] LNMi(-i) . . 015] TREBIRA -. 0041181 . 0016314 -2.5243[. 400] LNIP . 034486 . 040625 . 84888[. 28016 22985 229] LRUKYDS . . 1.2189[. 019253 016447 248] S1 -. . -1.1706[. 020343 015896 1. 2071 S2 . . -2798[. 997] S3 -. 5155E-4 . 011712 -. 0044010[. 998-73_ F-statistic 3[. 000] R-Squared . F( 7,47) 401 . 99808 S. E. 024536 R-Bar-Squared . of Regression . Residual Sum of Squares . 026296 Mean of Dependent Variable 10.5365 S. D. of Dependent Variable . 56005 Maximum of Log-likelihood 130.1988 DW-statistic 2.1156 Durbin's h-statistic -. 47440[. 635] Diagnostic Tests * Test Statistics * LM Version #F Version * A: Serial Correlation *CHI-SQ( 4)= 11.6429[. 00]*F( 4,43)= 2.8867[. 433]' * B: Functional Form *CHI-SQ( 1)= 1. i956[. 274]*F( 1,46)= 1.0222[. 317]) * C: Normality SCHI-SO( 2)= 2.5891C. 274]* Not applicable ***a *CHI-SO( 1)= 16650[. 16O93[. 69Q]y * D: Heteroscedasticity . 683]*F( 1,53)= . A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 32 Ordinary Least Squares Estimation Dependent variable is LNM1 55 observations used for estimation from 7502 to 8804 Regressor Coefficient Standard Error T-Ratio[Prob] CONST 7.4106 5.5543 1.3342[. 189] 1.3625 27687 LNM1(-1) . 4.9212[. 000] 0014730 TREBIRA -. . 0023136 -. 63665[. 5228] 14916 LNIP -. . 12237 -1.2189[. 2293 LRUKYDS -1.5631 1.1827 -1.3217[. 193] 015069 363] Si -. . 016400 -. 91885[. 019445 62 . . 015655 1.2421[. 221] 9009E-4 011527 S3 . . . 0078151[. 994] UHAUS 1.8340 1.1550 1.5879[. 119] R-Squared . 99842 F-statistic F( 8,46) 3624.80.000] 99814 S. E. R-Bar-Squared . of Regression . 024149 Sum Squares 026825 Mean Dependent Variable 10.5365 Residual of .. of S. D. of Dependent Variable . 56005 Maximum of Log-likelihood 131.6663 DW-statistic 2.0002 Durbin's h-statistic *NONE* Diagnostic Tests * Test Statistics * LM Version *F Version * * * * A: Serial Correlation *CHI-SO( 4)= 8.1040[. 088]*F( 4,42)= 1.8145[. 144]' * * * * BsFunctional Form *CHI-SO( i)= 3.0411[. 0813*F( 1,45)= 2.6338[. 112]' * * * * C; Normality *CHI-SO( 2)= 2.8935[. 235]* Not applicable * * * *CHI-SQ( 1)= 3O334[. 582)*F( * D: Heteroscedasticity . 1,53)= . 29393[. 590]) A: Lagrange multiplier test of residual serial correlation g: Samsey's RESET test using the square of the fitted values C: Sased on a test of skewness and kurtosis of residuals D: Based on the regression of squared residuals on squared fitted values TABLE 33 Ordinary Least Squares Estimation Dependent variable is LNM1 58 observations used for estimation from 7502 to 8903 Regressor Coefficient Standard Error T-Ratio[Prob] 20561 22167 92754[. 360] CONST . . . 007] LIPGNP 1.3342 . 46900 2.8448[. 046444 27678 16780[. 868] LRGGNP -. . -. GTRBR -. 011860 . 0063120 -1.8790[. 068] 15090 002] LNMI(-1) . 51404 . 3.4065[. 79230 088] LIPGNP(-1) -. . 45303 -1.7489[. 15623 21839 71539[. 4793 LRGGNP(-1) -. . -. 0064080 475] GTRBR(-1) -. . 0088798 -. 72164[. 22067 210] LNM1(-2) . . 17324 1.2738[. 016559 43240 038295[. 970] LIPGNP(-2) . . . 056192 800] LRGGNP(-2) -. . 22004 -. 25537[. 0010452 907] GTRBR(-2) . . 0088811 . 11769[. 036711 838] LNM1(-3) -. . 17786 -. 20640[. 53168 2353 LIPGNP(-3) -. . 44056 -1.2068[. 12903 67008[. 507] LRGGNP(-3) . . 19256 . 0068959 0095354 72319[. 474] GTRBR(-3) . . . 30819 17739 090] LNMI(-4) . . 1.7374[. 11103 LIPGNP(-4) -. . 48841 -. 22732[. 821] LRGGNP(-4) . 27014 . 20524 1.3162[. 196] GTRBR(-4) -. 0013906 . 0066269 -. 220984[. 835] R-Squared . 99777 F-statistic F(19,38) 896.1401[. 000] 99666 S. E. R-Bar-Squared . of Regression . 015225 0088080 Residual Sum of Squares . Mean of Dependent Variable 5.5718 26343 S. D. of Dependent Variable . Maximum of Log-likelihood 172.6851 DW-statistic 2.0552 Diagnostic Tests * Test Statistics * LM Version *F Version * A: Serial Correlation SCHI-SO( 4)= 7.7235[. 102]*F( 4,34)= 1.3058[. 288]* Form SCHI-SO( 1)= 78471[. 376]*F( 50746[. * B: Functional . 1,37)= . 481]* * C: Normality *CHI-SO( 2)= 1.3591[. 507]* Not applicable * D: Heteroscedasticity SCHI-SO( 1)= 2.0681[. 150]*F( 1,56)= 2.0706[. 156]4 * E: Predictive Failure *CHI-SO( 5)= 49.3080[. 000]*F( 5,38)= 9.8616C. 000]* A: Lagrange multiplie r test of residual serial correlation B: Ramsey's RESET tes t using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: Based on the regre ssion of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) TABLE 34 Variable Deletion Test (OLS case) Dependent variable is LNM1 List of the variables deleted from the regression: GTRBR(-1) LIPGNP(-1) LRGGNP(-1) LNM1(-2) GTRBR(-2) LIF'GNP(-2) LRGGNP(-2) LNM1(-3) GTRDR(-3) LIPGNP(-3) LRGGNP(-3) LNM1(-4) GTRBR(-4) LIPGNP(-4) LRGGNP(-4) 58 observations used for estimation from 7502 to 8903 Regressor Coefficient Standard Error T-Ratio[Prob] 38177 CONST -. . 19042 -2.0049[. 050] 011328 GTRBR -. . 0023204. -4.8818[. 000] 51855 000] LIPGNP . . 10113 5. "1278[. 1.1338 LRGGNF' . 12097 9.3726[. 000] 35908 LNM1(-1) . . 050294 7.1396[. 000] ***************##***##**#*#*****************#***********#*******************#*> Joint test of zero restrictions on the coefficient of deleted variables: Lagrange Multiplier Statistic CHI-SO(15)= 45.2796[. 000] Likelihood Ratio Statistic CHI-SQ(15)= 87.9996[. 000] F Stat. istic F(15,38)= 9.0177[. 000] TABLE 35 Variable Deletion Test (OLS case) Dependent variable is LNMI List of the variables deleted from the regression: LRGGNP LRGGNF(-1) GTRBR(-1) LIPGNP(-2) LRGGNP(-2) GTREBR(-2) LNM1(-3) LRGGNF(-3) GTRBR(-3) LIPGNP(-4) GTRBR(-4) 58 observations used for estimation from 7502 to 8903 **********#*****#***'*#*#***********#******************************************ý Regressor Coefficient Standard Error T-Ratio[Prob] 10915 13324 CONST . . . 81923[. 417] 1.1245 27209 LIPGNF . 4.1328[. 000] 013043 GTRBR -. . 0016601 =7.8567[. 000] 40071 LNM1(-i) . . 076563 5.2337[. 000] 70029 LIPGNF(-1) -. . 23469 -2.9839[. 004] 32381 LNMI(-2) . . 057716 5.6103[. 000] 36372 LIPGNP(-3) -. . 10305 -3.5294[. 001] 098076 LNMI(-4) . . 096766 1.0136[. 3.16] 46180 LRGGNP(-4) . . 13274 3.47900.0013 Joint test of zero restrictions on the coefficient of deleted variables: Lagrange Multiplier Statistic CHI-SO(11)= 8.5549[. 663] Likelihood Ratio Statistic CHI-SQ(11)= 9.2556[. 598] F(11,38)= F Statistic . 59770[. 819] TABLE 36 Ordinary Least Squares Estimation Dependent variable is LNM1 58 observations us ed for estimation from 7502 to 8903 Regressor Coefficient Standard Error T-Riatia[Prob] 10915 13324 CONST . . . 81923[. 417] 013043 0016601 GTRBR -. . -7.8567[. 000] 1.1245 LIPGNP . 27209 4.1328[. 000] 40071 LNM1(-1) . . 076563 5.23370.0001 70029 LIPGNP(-1) -. . 23469 -2.9839[. 004] 32381 LNM1(-2) . . 057716 5.6103[. 000] 36372 LIPGNP(-3) -. . 10305 -3.5294[. 001] 098078 LNM1(-4) . . 096766 1.0136[. 316] 46180 13274 LRGGNP(-4) . . 3.4790[. 001] ' R-Squared . 99739 F-statistic F( 8, 49) 2338.7[. 000] 99696 S. E. Regression R-Bar-Squared . of . 014521 Residual Sum of Sq uares . 010332 Mean of Dependent Variable 5.5718 S. D. of Dependent Variable . 26343 Maximum of Log-lik elihood 168.0573 DW-statistic 1.8432 Diagnostic Tests * Test Statistics # LM Version *F Version *** ýI Serial Correlation SCHI-SO( 4)= 2.8819[. 5783*F( 4,45)= 58822E. 67311 * A: . * E3: Functional Form *CHI-SQ( i)= 1.2390[. 266]*F( 1,48)= 1.0478[. 311]1 ***I * C: Normality *CHI-SO( 2)= 1.5709[. 456]ß Not applicable I *** ýI *CHI-SQ( 1)= ('23815[. 877]*F( D: Heteroscedasticity . 1,56)= . o23004[. oac']I *** ýI * E: Fredictive Failure *CHI-SQ( 5)= 101.1438[. 0003*F( 5,49)= 20.2288[. 000]I #* * *********#**********************************IC****#**#**************** *ýI A: Lagrange multiplier test of residual serial correlation B: Ramsey's RESET test using the square of the fitted values C: Based on a test of skewness and kurtosis of residuals D: E+ased on the regression of squared residuals on squared fitted values E: A test of adequacy of predictions (Chow's second test) 1)A'1A: u, 085. IP UKYDS M1 BACREIN BACROIN 7401 27.1000 18494.0 12123.0 10655.0 16721.0 7402 28.0000 20140.0 12520.0 11415.0 17000.0 7403 30.2000 22155.0 12798.0 12638.0 17804.0 7404 32.0000 23890.0 13714.0 13074.0 18180.0 7501 34.0000 23845.0 13763.0 13617.0 18710.0 7 502 36.6000 25820.0 15292.0 "10757.0 18465.0 7503 38.6000 27502.0 16032.0 10825.0 19133.0 7504 39.8000 29816.0 16680.0 10600.0 19543.0 7601 41.3000 29748.0 16906.0 10454.0 19612.0 7602 42.3000 30377.0 17468.0 10679.0 20371.0 7603 43.5000 32532.0 18560.0 11495.0 21182.0 7604 44.8000 35125.0 18897.0 12490.0 22423.0 7701 46.7000 33804.0 18392.0 12799.0 22385.0 7702 47.6000 35423.0 19825.0 12849.0 23108.0 7703 48.9000 37549.0 20796.0 13790.0 23820.0 7704 49.8000 40344.0 22695.0 13857.0 24247.0 7801 52.8000 39448.0 22886.0 14365.0 24612.0 7802 53.6000 41116.0 23594.0 14812.0 26051.0 7803 54.7000 43306.0 24621.0 15408.0 26047.0 7804 55.3000 45753.0 26137.0 15651.0 26931.0 7901 57.7000 44027.0 25659.0 16488.0 27741.0 7902 59.2000 47813.0 26590.0 17301.0 28619.0 79Q3 62.3000 51815.0 27376.0 18220.0 29270.0 7904 64.5000 54803.0 28318.0 18783.0 30758.0 8001 60.1000 54096.0 27188.0 19872.0 31714.0 8002 70.5000 55511.0 27595.0 20962.0 33036.0 8003 73.8000 60264.0 28388.0 23297.0 35177.0 8004 76.5000 62676.0 30161.0 22960.0 35901.0 8101 78.3000 61209.0 29398.0 22908.0 37255.0 8102 78.5000 62077.0 31077.0 22813.0 39314.0 8103 80.1000 65165.0 31727.0 24710.0 43133.0 8104 81.3000 67915.0 34859.0 24594.0 47211.0 8201 82.7000 66963.0 34244.0 26418.0 53936.0 8202 84.4000 67554.0 36777.0 26828.0 58192.0 8203 85.9000 70778.0 37819.0 27984.0 61844.0 8204 87.1000 74286.0 40035.0 29047.0 65805.0 8301 88.7000 73668.0 41290.0 29562.0 71672.0 8302 88.7000 72982.0 42435.0 28915.0 73419.0 8,0 90.6000 77496.0 42766.0 30479.0 77060.0 8304 91.6000 81276.0 44539.0 31433.0 82238.0 8401 92.7000 78819.0 46372.0 32349.0 86100.0 8402 94.4000 77957.0 48499.0 33392.0 90684.0 8403 95.0000 81228.0 49855.0 34035.0 95006.0 8404 96.9000 86627.0 51558.0 36360.0 102955.0 8501 98.4000 85258.0 53462.0 38362.0 112767.0 8502 99.5000 85685.0 55947.0 36406.0 114522.0 850.3 100.4000 89767.0 58507.0 36482.0 114927.0 8504 101.6000 95229.0 60957.0 36382.0 118697.0 8601 101.6000 91570.0 64191.0 36861.0 122277.0 8602 102.2000 92113.0 68370.0 37831.0 129989.0 OBS. IP UKYDS M1 BACREIN BACROIN 8603 102.8000 96772.0 73307.0 38429.0 137682.0 8604 103.8000 102680.0 74695.0 39612.0 149833.0 8701 105.3000 99244.0 79170.0 39893.0 157659.0 8702 107.6000 100624.0 84982.0 42301.0 171342.0 6703 108.6000 107250.0 88664.0 41571.0 180908.0 6704 109.7000 113739.0 91866.0 41853.0 190749.0 8801 111.3000 109690.0 95799.0 42767.0 196187.0 8802 113.6000 111742.0 100883.0 45595.0 206869.0 8803 116.0000 119512.0 103987.0 52258.0 221554.0 8804 118.1000 126289.0 105048.0 55801.0 233364.0 8901 121.1000 122509.0 *NONE* 58151.0 250551.0 8902 123.4000 123321.0 *NONE* 62189.0 267262.0 8903 124.5000 129580.0 *NONE* 69420.0 3092287.0 8904 125.4000 136089.0 *NONE* 71480.0 324492.0 9001 126.7000 132339.0 *NONE* 73250.0 332208.0 9002 133.0000 133459.0 *NONE* 74957.0 342310.0 9003 136.5000 138845.0 *NONE* 74814.0 346054.0 9004 136.9000 144869.0 *NONE* 76972.0 356647.0 9101 139.5000 136597.0 *NONE* 76913.0 364411.0 9102 141.9000 140385.0 *NONE* 76976.0 373765.0 9103 143.7000 146129.0 *NONE* 75280.0 375420.0 9104 144.0000 152253.0 *NONE* 74870.0 384545.0 9201 145.1000 *NONE* *NONE* 74471.0 359764.0 9202 147.5000 *NONE* *NONE* 72256.0 360748.0 9207, 148.1000 *NONE* *NONE* 71540.0 358591.0 9204 *NONE* *NONE* *NONE* *NONE* *NONE* OBS. TREBIRA 80V RLCB LNM1 LBACROIN LBACREIN 7401 *NONE* *NONE* *NONE* 9.4029 9.7244 9.2738 7402 11.2400 13.0000 9.5000 9.4351 9.7410 9.3427 7403 10.9800 13.0000 9.5000 9.4570 9.7872 9.4445 7404 10.9900 13.0000 9.5000 9.5262 9.8081 9.4784 7501 9.3700 11.5000 7.5000 9.5297 9.8368 9.5191 7502 9.4800 10.5000 6.2500 9.6351 9.8236 9.2833 75Q3 10.4800 11.0000 6.5000 9.6823 9.8592 9.2896 7504 10.6400 12.0000 7.0000 9.7220 9.8804 9.2686 7601 8.4200 10.5000 5.5000 9.7354 9.8839 9.2547 7602 10.9900 11.5000 6.5000 9.7681 9.9219 9.2760 7603 12.3500 13.0000 8.5000 9.8288 9.9609 9.3497 7604 13.5100 15.0000 11.0000 9.8468 10.0178 9.4327 77Q1 9.3500 10.5000 5.0000 9.8197 10.0161 9.4571 7702 7.4600 9.5000 4.0000 9.8947 10.0479 9.4610 77Q3 5.3000 8.0000 3.0000 9.9425 10.0783 9.5317 7704 6.2900 8.1200 4.0000 10.0299 10.0960 9.5365 7801 5.9900 7.5000 3.0000 10.0383 10.1110 9.5725 7802 9.2700 11.0000 6.7500 10.0687 10.1678 9.6032 7803 9.1700 11.0000 6.7500 10.1114 10.1677 9.6426 7804 11.5600 13.5000 10.0000 10.1711 10.2010 9.6583 7901 11.4400 14.0000 10.5000 10.1526 10.2307 9.7104 7902 13.3300 15.0000 11.5000 10.1883 10.2618 9.7585 79Q3 13.3600 15.0000 11.5000 10.2174 10.2843 9.8103 7904 15.8400 18.0000 15.0000 10.2513 10.3339 9.8407 8001 16.2800 18.0000 15.0000 10.2105 10.3645 9.8971 8002 15.6800 18.0000 15.0000 10.2254 10.4054 9.9505 8003 14.4000 17.0000 14.0000 10.2537 10.4681 10.0561 8004 13.0700 15.0000 11.6800 10.3143 10.4885 10.0415 8101 11.5300 13.6400 9.9100 10.2887 10.5255 10.0392 81Q2 12.0900 13.0000 9.0000 10.3442 10.5793 10.0351 8103 13.9600 14.0000 10.2500 10.3649 10.6720 10.1150 8104 14.5100 15.6000 12.5500 10.4591 10.7624 10.1103 8201 12.4900 14.2200 10.5400 10.4413 10.8956 10.1818 8202 12.2300 13.6400 9.6900 10.5126 10.9715 10.1972 82Q3 9.9100 11.5000 7.3100 10.5406 11.0324 10.2394 82Q4 9.9000 11.0600 6.7500 10.5975 11.0945 10.2767 8301 10.4100 11.7200 7.7200 10.6284 11.1799 10.2942 83Q2 9.4700 10.7300 6.3400 10.6557 11.2039 10.2721 6303 9.1600 10.5000 6.0000 10.6635 11.2523 10.3248 8304 6.8700 10.0000 5.5000 10.7041 11.3174 10.3556 8401 8.4300 9.7400 5.3100 10.7445 11.3633 10.3843 8402 8.8600 10.1500 5.8100 10.7893 11.4151 10.4161 8403 10.0200 11.5000 7.3100 10. B169 11.4617 10.4351 8404 9.1000 10.5600 6.2500 10.8505 11.5420 10.5012 8501 12.9300 14.8000 10.1300 10.8867 11.6331 10.5548 6502 11.8900 13.5500 9.3500 10.9322 11.6485 10.5025 8503 11.0600 12.5000 7.8600 10.9769 11.6521 10.5046 6504 11.1500 12.5000 7.8600 11.0179 11.6843 10.5018 8601 11.0600 13.1300 8.3500 11.0696 11.7140 10.5149 8602 9.3200 11.0000 6.0800 11.1327 11.7752 10.5409 OBS. TREBIRA ROV RLCB LNM1 LBACROIN LBACREIN 8603 9.6100 11.0000 6.0800 11.2024 11.8327 10.5566 8604 10.6600 12.0000 6.9200 11.2212 11.9173 10.5869 8701 9.3500 11.4300 6.1600 11.2794 11.9682 10.5940 8702 8.5400 10.0000 4.6900 11.3502 12.0514 10.6526 8703 9.6900 11.0000 4.9000 11.3926 12.1057 10.6352 8704 8.1900 8.5700 3.6700 11.4281 12.1587 10.6419 8801 8.4400 8.7600 3.8200 11.4700 12.1868 10.6635 8802 8.4900 8.5700 2.8700 11.5217 12.2398 10.7276 88033 11.5600 13.0000 5.1200 11.5520 12.3084 10.8639 8604 12.5500 14.0000 5.7000 11.5622 12.3604 10.9295 8901 12.4100 14.0000 5.9700 *NONE* 12.4314 10.9708 8902 13.5900 15.0000 5.9700 *NONE* 12.4960 11.0379 8903 13.4400 15.0000 5.9700 *NONE* 12.6420 11.1479 8904 14.5000 16.0000 6.5900 *NONE* 12.6900 11.1772 9001 14.5900 16.0000 6.5900 *NONE* 12.7135 11.2016 9002 14.3200 16.0000 6.5900 *NONE* 12.7435 11.2247 9003 14.2300 16.0000 6.5900 *NONE* 12.7544 11.2228 9004 12.1100 15.0000 5.1100 *NONE* 12.7845 11.2512 9101 11.5600 14.0000 5.1100 *NONE* 12.8060 11.2504 9102 10.7500 12.5000 5.1100 *NONE* 12.8314 11.2512 9103 9.6600 11.5500 *NONE* *NONE* '12.8358 , 11.2290 9104 10.1000 11.5000 *NONE* *NONE* 12.8598 11.2235 9201 10.1000 11.5000 *NONE* *NONE* 12.7932 11.2182 9202 9.4200 11.0000 *NONE* *NONE* 12.7959 11.1880 9203. 9.1600 9.7700 *NONE* *NONE* 12.7899 11.1780 9204 *NONE* *NONE* *NONE* *NONE* *NONE* *NONE* O6S. LBACRTO LRUKYDS 7401 10.2174 6.5257 7402 10.2547 6.5783 7403 10.3236 6.5980 7404 10.3499 6.6155 7501 10.3837 6.5297 7502 10.2827 6.5589 7503 10.3076 6.5688 7504 10.3137 6.6189 7601 10.3112 6.5797 7602 10.3434 6.5767 7603 10.3944 6.6172 7604 10.4606 6.6645 7701 10.4683 6.5846 7702 10.4901 6.6123 7703 10.5350 6.6436 7704 10.5481 6.6972 7801 10.5707 6.6162 7802 10.6180 6.6426 7803 10.6324 6.6742 7804 10.6592 6.7182 7901 10.6971 6.6373 7902 10.7347 6.6941 7903 10.7683 6.7235 7904 10.8106 6.7448 6001 10.8510 6.6775 8002 10.8967 6.6687 8003 10.9763 6.7051 8004 10.9829 6.7084 8101 11.0048 6.6615 8102 11.0369 6.6730 8103 11.1250 6.7014 8104 11.1817 6.7279 8201 11.2942 6.6967 8202 11.3506 6.6851 82Q3 11.4057 6.7141 8204 11.4601 6.7486 8301 11.5252 6.7221 8302 11.5360 6.7127 8303 11.5856 6.7515 8304 11.6411 6.7882 8401 11.6822 6.7455 8402 11.7286 6.7164 840-7 11.7679 6.7511 8404 11.8445 6.7957 8501 11.9259 6.7644 8502 11.9246 6.7583, 850.3 11.9277 6.7958 8504 11.9517 6.8430 8601 11.9775 6.8038 8602 12.0306 6.8038 OBS. LBACRTO LRUKYDS 8603 12.0789 6.8473 8604 12.1519 6.8969 6701 12.1938 6.8485 8702 12.2721 6.8407 8703 12.3126 6.8952 8704 12.3571 6.9439 8801 12.3840 6.8932 8802 12.4390 6.8913 6803 12.5202 6.9376 8804 12.5748 6.9748 8901 12.6401 6.9193 8902 12.7052 6.9071 8903 12.8445 6.9477 8904 12.8891 6.9896 9001 12.9128 6.9356 9002 12.9415 6.9112 9003 12.9501 6.9248 9004 12.9799 6.9643 9101 12.9975 6.8867 9102 13.0186 6.8970 9103 13.0186 6.9245 9104 13.0377 6.9635 9201 12.9813 *NONE* 9202 12.9785 *NONE* 9203 12.9718 *NONE* 9204 *NONE* *NONE* VATA: 651204Njo CBS. M1 IPGNP GGNP BACREMA GTRBR RCRCA 7401 132.9000 63.6000 229.1000 114.1000 7.1200 *NONE* 7402 140.8000 64.8000 238.9000 117.7000 5.7100 *NONE* 7403 141.5000 65.9000 250.8000 119.0000 5.7100 *NONE* 7404 158.4000 69.2000 264.9000 120.6000 5.1900 *NONE* 7501 149.3000 68.2000 237.8000 117.1000 3.3800 11.5000 7502 160.6000 69.1000 249.2000 115.3000 3.3800 10.2800 7503 164.4000 69.2000 259.7000 112.7000 3.1300 9.0900 7504 179.9000 72.3000 281.0000 114.7000 3.1300 8.8500 7601 166.8000 70.5000 260.7000 110.0000 3.1300 8.6800 7602 180.2000 71.6000 274.7000 112.9000 3.1300 6.3400 7603 176.9000 72.4000 283.0000 114.1000 3.1800 8.3400 7604 186.9000 74.4000 305.4000 119.2000 3.1800 8.3200 7701 179.7000 72.9000 280.2000 116.6000 3.1800 8.2600 7702 190.5000 74.4000 291.1000 119.2000 3.1800 7.9600 7703 193.1000 74.7000 296.9000 119.2000 3.1800 7.8300 7704 208.1000 77.7000 327.4000 125.2000 2.6700 7.7400 7801 204.2000 76.1000 300.0000 118.8000 2.6700 7.3400 7802 215.4000 77.4000 314.7000 122.4000 2.6700 7.2700 7803 217.5000 78.4000 324.2000 123.0000 2.6700 7.3000 7804 237.9000 80.5000 350.5000 128.6000 2.6700 7.2900 7901 225.5000 79.1000 322.4000 127.4000 3.6800 7.4200 7902 233.1000 80.0000 340.4000 133.1000 3.6800 8.2700 7903 230.2000 81.4000 351.2000 134.9000 4.7000 9.2400 7904 247.9000 84.2000 379.8000 144.0000 5.7300 10.3900 8001 228.7000 82.7000 352.4000 138.5000 6.7600 11.6600 8002 237.1000 84.5000 361.8000 144.0000 7.2800 12.5400 8003 237.8000 85.4000 369.0000 142.2000 7.2800 12.5300 8004 257.3000 88.1000 394.2000 174.0000 7.2800 12.5700 8101 232.4000 85.9000 362.6000 175.6000 7.2800 14.2100 8102 242.5000 87.5000 373.8000 178.9000 7.2800 15.2900 8103 234.3000 88.7000 386.1000 178.1000 7.2800 15.4400 8104 255.3000 92.5000 417.1000 181.4000 7.2800 15.0100 8201 237.6000 90.2000 377.9000 175.8000 7.2800 14.6100 8202 250.7000 91.4000 389.0000 177.6000 7.2800 13.6100 8203 248.5000 92.8000 397.0000 179.1000 6.7600 13.0600 8204 273.1000 96.1000 426.4000 180.8000 4.7000 11.1300 8301 263.4000 93.6000 395.0000 173.1000 3.6800 10.7200 8302 277.9000 94.4000 407.5000 176.1000 3.6800 9.7700 8303 274.0000 95.9000 417.4000 177.1000 3.6800 9.8000 8304 295.8000 99.2000 455.8000 181.9000 3.6800 9.7700 8401 272.5000 96.2000 421.8000 178.8000 3.6800 9.7800 8402 282.8000 96.6000 422.9000 181.7000 4.1900 9.7800 8403 281.5000 97.3000 439.6000 182.50010 4.1900 9.8900 8404 314.2000 101.1000 479.0000 186.0000 4.1900 9.7800 8501 285.2000 98.0000 428.8000 188.9000 4.1900 9.8000 8502 294.4000 98.4000 444.0000 191.9000 4.1900 9.7500 8503 297.8000 99.7000 463.3000 188.3000 3.6800 9.1400 8504 334.1000 103.6000 498.4000 197.4000 3.6800 9.1300 313.4000 100.9000 447.5000 <3601 199.4000 3.1800 8.8400 329.3000 101.9000 474.0000 8602 "202.5000 3.1800 8.6900 O8S. M1 IPGNP GGNP BACREMA GTRBR RCRCA 8603 326.9000 103.1000 488.6000 201.5000 3.1800 8.6400 8604 358.8000 106.9000 526.0000 206.2000 3.1800 8.6300 8701 336.8000 103.6000 467.6000 203.7000 2.6700 8.4500 8702 358.7000 104.3000 488.3000 204.4000 2.6700 8.2900 8703 357.2000 104.4000 502.3000 208.0000 2.6700 8.2800 8704 385.2000 108.5000 544.8000 208.9000 2.1600 8.1800 8801 369.5000 104.7000 495.3000 208.0000 2.1600 8.0700 8802 393.5000 105.9000 511.9000 212.9000 2.6700 8.0600 8803 389.1000 106.3000 528.7000 213.9000 3.1800 8.7000 8804 427.0000 110.4000 572.1000 221.1000 3.1800 8.6900 8901 403.2000 107.4000 530.6000 225.1000 3.6900 9.2600 8902 412.0000 108.4000 548.8000 231.6000 4.7000 9.7300 8903 408.7000 109.1000 558.8000 237.1000 4.7000 10.1400 8904 450.6000 113.7000 607.0000 243.5000 5.7300 11.1000 9001 412.8000 110.8000 570.3000 248.8000 5.7300 11.5200 90Q2 483.2000 112.0000 585.9000 254.4000 5.7300 11.6000 9003 502.8000 113.4000 612.6000 257.4000 5.7300 11.6900 9004 584.2000 117.2000 656.7000 263.0000 5.7300 11.9700 9101 530.4000 *NONE* *NONE* 343.4000 6.2500 12.2200 91Q2 541.0000 *NONE* *NONE* 354.5000 6.2600 12.2600 9103 546.9000 *NONE* *NONE* 357.9000 7.2800 12.9200 9104 604.3000 *NONE* *NONE* 359.9000 7.8000 12.9500 9201 556.6000 *NONE* *NONE* 356.0000 *NONE* 13.3800 9202 576.4000 *NONE* *NONE* 364.2000 *NONE* 13.4400 9203 588.2000 *NONE* *NONE* 359.4000 *NONE* 14.0500 9204 669.6000 *NONE* *NONE* 339.9000 *NONE* *NONE* OBS. LNMI LRGGNP LIPGNP LBACREMA 7401 4.6896 1.2815 4.1526 4.7371 7402 4.9473 1.3047 4.1713 4.7681 7403 4.9523 1.3365 4.1881 4.7791 7404 5.0651 1.3424 4.2370 4.7925 7501 5.0060 1.2490 4.2224 4.7630 7502 5.0789 1.2827 4.2356 4.7475 7503 5.1023 1.3225 4.2370 4.7247 7504 5.1924 1.3575 4.2808 4.7423 7601 5.1168 1.3078 4.2556 4.7005 7602 5.1941 1.3446 4.2711 4.7265 7603 5.1756 1.3632 4.2822 4.7371 7604 5.2306 1.4122 4.3095 4.7808 7701 5.1913 1.3464 4.2891 4.7587 7702 5.2497 1.3642 4.3095 4.7808 7703 5.2632 1.3799 4.3135 4.7808 7704 5.3380 1.4383 4.3529 4.8299 78Q1 5.3191 1.3717 4.3320 4.7774 7802 5.3725 1.4026 4.3490 4.8073 7803 5.3822 1.4195 4.3618 4.8122 7804 5.4719 1.4711 4.3883 4.8567 7901 5.4183 1.4051 4.3707 4.8473 7902 5.4515 1.4481 4.3820 4.8911 7903 5.4389 1.4620 4.3994 4.9045 7904 5.5130 1.5064 4.4332 4.9698 8001 5.4324 1.4495 4.4152 4.9309 8002 5.4685 1.4543 4.4368 4.9698 8003 5.4714 1.4635 4.4473 4.9572 8004 5.5502 1.4984 4.4785 5.1591 8101 5.4485 1.4401 4.4532 5.1682 8102 5.4910 1.4521 4.4716 5.1868 8103 5.4566 1.4708 4.4853 5.1823 8104 5.5424 1.5061 4.5272 5.2007 8201 5.4706 1.4326 4.5020 5.1693 8202 5.5243 1.4483 4.5152 5.1795 8203 5.5154 1.4535 4.5304 5.1879 8204 5.6098 1.4900 4.5654 5.1974 8301 5.5737 1.4399 4.5390 5.1539 8302 5.6273 1.4625 4.5475 5.1711 8303 5.6131 1.4707 4.5633 5.1767 8304 5.6897 1.5249 4.5971 5.2035 8401 5.6076 1.4781 4.5664 5.1863 8402 5.6447 1.4766 4.5706 5.2024 8403 5.6401 1.5081 4.5778 5.2068 8404 5.7500 1.5556 4.6161 5.2257 8501 5.6532 1.4760 4.5850 5.2412 8502 5.6849 1.5068 4.5890 5.2570 8503 5.6964 1.5362 4.6022 5.2380 8504 5.8114 1.5709 4.6405 5.2852 8601 5.7475 1.4895 4.6141 5.2953 8602 5.7970 1.5372 4.6240 5.3107 OBS. LNM1 LRGGNP LIPGNP LBACREMA 8603 5.7897 1.5558 4.6357 5.3058 8604 5.8828 1.5934 4.6719 5.3288 8701 5.8195 1.5071 4.6405 5.3166 8702 5.8825 1.5437 4.6473 5.3201 8703 5.8783 1.5710 4.6482 5.3375 8704 5.9538 1.6137 4.6868 5.3419 8801 5.9122 1.5541 4.6511 5.3375 8802 5.9751 1.5756 4.6625 5.3608 8803 5.9638 1.6042 4.6663 5.3655 8804 6.0568 1.6452 4.7041 5.3986 8901 5.9994 1.5974 4.6766 5.4165 8902 6.0210 1.6219 4.6858 5.4450 8903 6.0130 1.6335 4.6923 5.4685 8904 6.1106 1.6750 4.7336 5.4951 9001 6.0230 1.6384 4.7077 5.5166 90Q2 6.1804 1.6547 4.7185 5.5389 9003 6.2202 1.6868 4.7309 5.5506 9004 6.3702 1.7233 4.7639 5.5722 9101 6.2736 *NONE* *NONE* 5.8389 91Q2 6.2934 *NONE* *NONE* 5.8707 9103 6.3043 *NONE* *NONE* 5.8803 9104 6.4041 *NONE* *NONE* 5.8858 92Q1 6.3218 *NONE* *NONE* 5.8749 9202 6.3568 *NONE* *NONE* 5.8977 9203 6.3771 *NONE* *NONE* 5.8844 9204 6.5067 *NONE* *NONE* 5.8287 104 CHAPTER 4 IS IT MONEY, CREDIT OR BOTH, OR NEITHER? AN EMPIRICAL INVESTIGATION BASED ON THE FREE LIQUIDITY RATIO FOR COMMERCIAL BANKS. This chapter is based on my paper "Free Liquidity Ratio for Commercial Banks: an Estimate with Italian Data" published (in English) on the Italian review "Giornale degli Economists e Annali di Economia", vol. L, pp. 399-424, July-August 1991. I am very Wiji grateful to Giovanni'Amisano, Narendranathan, and the anonymous referees of the above-mentioned review for their helpful comments. I to Norman Ireland am also indebted and Keith Cowling for their None them is valuable suggestions. of responsible for any mistake that might be found here 105 IS IT MONEY, CREDIT OR BOTH, OR NEITHER? AN EMPIRICAL INVESTIGATION BASED ON THE FREE LIQUIDITY RATIO FOR COMMERCIAL BANKS. 1. Introduction The previous chapters have analyzed the macroeconomic implications of oligopsony in the market for credit, and the implications of financial systems' securitization. Since some emphasis has been put on the behaviour of the banking system, it would be interesting, at this point, to have a closer look at the assumptions concerning the behaviour of the banks. In particular, in accordance'with Bernanke and Blinder's [1988] model - signalled in chapter 3 as an important theoretical foundation for the "institutional analysis" performed there - the banking system has been assumed to play an important role in the allocation of the financial funds, as argued by the "credit view", in contrast with the traditional orthodox approach that described the banking system "veil" as a mere passive on the economy. The main differences between the two approaches lie, of course, in the relevance of information market imperfections and asymmetries, but also, as a in the to consequence, extent which the 'autonomous financial investment decisions of the banks affect the aggregate level of investment. If the physical non-remunerated liquid assets in the banks' portfolios reflect a non-investment decision, a decision to investment decision, then postpone an a detailed empirical analysis banks' "liquidity of the preference" could turn out to be an interesting- although only partial - test on the reliability of a few assumptions of the "credit view". 106 The commercial banks' free liquid reserves have been the object of a few empirical studies in the 1970s and in the early 1980s, in connection with the debate on the stability of the demand for money (for instance Cagan [1969], Bryant [1983], Langhor [1981], Richter and Teigen [1982], Wessels [1982]). Surprisingly, such empirical studies do not seem any longer to be a very central topic, although the free liquid reserves play an important theoretical role for at least two reasons. First of all, the free liquidity ratio contributes to determine the money multiplier, and, therefore, its behaviour directly affects the behaviour of the money stock. Secondly, the free liquid reserves of commercial banks are commonly regarded as a "shock absorber", due to their high degree of liquidity. This implies that, if the bank's willingness to supply funds determines the aggregate level of credit and liquidity, then - given the balance sheet equality between banks' deposits and banks' level free liquid assets - the of reserves must be negatively correlated to the value of less liquid financial assets. The next section briefly surveys the empirical literature on commercial banks' free reserves and its theoretical foundations. Section 3 contains a brief non-formal description of the literature irreversible investments on (partially) under uncertainty, and a few how considerations on this kind of literature might be used as an interpretative tool to describe the behaviour of free liquid 4 reserves. Section presents a few empirical estimates based on a incorporates into the model that standard approach - based on the theory portfolio allocation -a few theoretical aspects of the in 3. literature analyzed section Section 5 draws a few conclusions. be In particular, it will argued that the empirical results shown in 107 this paper are consistent with the "credit view", such as illustrated in the famous contribution by Bernanke and Blinder [1988]. 2. A few comments on some standard empirical works. The behaviour of the free liquidity ratio for commercial banks has often been related to the availability of liquidity and borrowing from the central bank (Langhor [1981], Wessels [1982], Richter and Teigen [1982]). The traditional background is provided by the portfolio allocation theory. The explanatory variables usually considered are the level of liquidity, such as determined by the public and by the foreign sector, some kind of opportunity cost, and the own yields on liquid assets, all of which summarized by some relevant interest rate. In most empirical contributions, as often happens for the estimates of financial variables, the explanatory describing borrowing bank, power of the equations from the central banks' free liquidity is high. or commercial ratio not very This fact is often justified by the high volatility of financial variables. The free reserves of commercial banks are in general assumed to depend positively on their own yield, and negatively on the yield on in alternative assets, accordance with the standard wealth In the free allocation theory. addition, liquid reserves can be "shock for regarded as a sort of absorber" commercial banks. In such interpretation the a context, a new of free liquidity ratio is in the possible, as we will see next section. 108 The rate of interest on riskless alternative liquid assets in Italy is traditionally the rate on short term treasury bills. The own yield on free reserves in Italy has been fixed at 0.5% for more than thirty years: obviously such a variable cannot describe properly the dynamics of the own yield on banks' free reserves. Arguing that free reserves are most required the more frequently the required reserves fluctuate, Richter and Teigen [1982] use the variance of constrained reserves as a proxy variable for the yield on free reserves. Since the required reserves ratio has been subject in Italy only to rare modifications, this solution does not seem to in apply to the Italian case, and any case, the variance of liquid constrained reserves over twelve months turned out to be non significant in preliminary analyses performed with our data. Some early studies (for example Cagan [1969]) consider the rate from for of interest on banks' borrowing the central bank as a proxy the yield on free reserves, because the higher the free reserves is ratio, the less likely it that the banks need to borrow from the central bank; and the higher the interest rate on banks' borrowing, the more convenient it is to keep reserves. This point could be criticized because the same reasoning could apply to the interbank "call rate, or to the money rate", whose behaviour is closely related to that of the treasury bill rate, which is regarded as an opportunity cost and not as a yield on reserves. To the extent that the free reserves of commercial banks play the role of "shock absorbers", an empirical specification describing their behaviour should include the information about shocks in the determined by supply of liquidity the public and foreign sectors. For this reason, Richter and Teigen [1982] included two more 109 liquid banks in their variables for free reserves of commercial between the equation: the first one is the ratio the variation of deposits. The net central bank balances of the state, and the total foreign balances second is the ratio of the variation of the net of the central bank and the total deposits. Richter and Teigen point out that the signs of the coefficients referred to these two in regressors cannot be established a priori, and indeed their latter estimates the sign of the coefficient of the variable changes throughout two different sample periods. In fact their size and dynamic path might reflect or determine expectations about future if, for foreign policy interventions: example, the net position of the central bank showed a persistent drastic variation, one could bank, the expect some intervention of the central meant to affect liquidity of the system, according to the monetary policy targets. Similarly, the size and dynamic path of the net central bank balances of the state could determine the expectations of some other policy decisions, such as open market operations. All these reasons by suggest that the dynamic path of the liquidity created the state be and foreign sectors may relevant not only for its direct effect, but also because they contribute to determine expectations about policy interventions. The rate of inflation, and, possibly, its rate of growth, could capture some of the opportunity costs of holding free liquid in reserves, especially a context like the Italian one, where the has been kept yield on free reserves constant at a very low level for Gthe* g time. setting of our model is only partly similar to that off Richter and Teigen [1982], because the own yield on free liquid 110 reserves has been interpreted here according to the above-mentioned literature on investments with sunk costs under uncertainty. The next section contains a brief explanation of such theoretical approaches. 3. A new possible interpretation of the free liquidity ratio for commercial banks. A new tool of interpretation for the free liquidity ratio of commercial banks may be provided by the recent literature on investment decisions under conditions of irreversibility and uncertainty (Bertola and Caballero [1991], Dixit [1992a], [1992b], Pindyck [1991]) - which provides a useful framework of interpretation for the decision of "non-investing" - and by a recent "new-keynesian" contribution which applies the conceptual framework of irreversible investment under uncertainty to the case where the decision of "non-investing" corresponds to a decision of investing in liquid assets (Chamley [1993]). The purpose of the discussion of the present section is to provide an alternative theoretical interpretation of the variable "yield" on the free reserves of commercial banks which has always been a weak point of the empirical works on this issue, given the fact that, as is well known, free reserves are usually non-remunerated and, therefore, the yield justifying their existence has to be defined on a more theoretical ground. Since many of the contributions mentioned early in this section provide exhaustive surveys on the literature on investments with sunk costs under uncertainty, only a brief summary of the main points of this approach will be presented in what follows. 111 The starting point of the literature on irreversible investments under uncertainty is provided by the observation that implicit most models of investment decision are based on the is assumption that the investment expenditure is reversible. This also one of the basic implicit hypotheses of the familiar "net present value" rule, stating that firms should take up a project of investment when the net present value of its cash flow is at least as large as its costs. In the real world, on the other hand, the investment decision usually takes place under the following conditions: a) sunk costs, due to the fact that the investment might be firm-specific or industry-specific; in the case of bank loans and financial investments the presence of sunk costs - from the point of view of the bank - is due to the presence of "lemon problems": in 3' other words, the bank invests in financial assets carrying some intrinsic risk of capital losses, due not only to the possible bankruptcy of a borrower, but also to potential unexpected changes in the market interest rates, susceptible to determine a change in the market value of the bank loans and other financial assets; ýI b) "on-going" uncertainty concerning the future profitability of the specific investment, the latter being only inferred on the basis of some probability calculus; c) relevance of decision timing: in other words, the investment can be delayed, allowing the firm (in our case the bank) to collect further information on all of the variables affecting the investment before its profitability, committing resources. 112 The three above-mentioned conditions imply that the investor (i. e., in this case the bank) has to take into account the presence of a potential positive "value of waiting". In this context, the investment decision has been described by using the parallel of the "call option" in financial economics: the financial investment is not undertaken 'when the financial investment is "only just in the money", while it is undertaken when the option is "well in the money". In other words, the traditional "net present value" rule could be transformed into a rule suggesting that an investment should be undertaken if the net present value of its cash flow exceeds the purchase and installation cost by an amount at least equal to the value of keeping the option to invest the same resources elsewhere. The problem could be simply described as follows. Let xt be a variable directly affecting the level of financial profits (for instance, the price of the financial asset under consideration), let n(xt) be an indicator of the profitability of the financial investment, and let F(xt+llxt) be the probability distribution of xt+ltxt, i. e. The probability distribution of xt+l given the value of xt. Then, the value of having in hand a financial investment can be defined as follows: co V1(x*) = Eo C ptn(xt)Ix0=x* ) t=O where pt is the factor of discount from the (future) time "t" to the present time to and Ep stands for "expectation at time to. The value Vp of having the option is given by the value of being free between choosing to invest immediately (by paying a given 113 immediate cost, say "k") and postponing the investment decision by one period. Formally, this could be described as follows. VO = max ( Vl(xt) - k, R"E[VO(xt+l)Ixt]} In the presence of uncertainty, and for given expectations of for investor xt+l, it may be rational the to postpone the investment and wait for new information. It can be proved12 that the higher the level of "on-going" uncertainty, the higher the value of "keeping invest in alive" the option to the future relative to the value of "having in hand" the investment project. In other words, an increase in the level of "on-going" uncertainty increases the "value of information". Dixit (1989,1992a, 1992b) has employed this simple idea to formally explain some empirical phenomena, such as the high persistence of relatively profit margins in the absence of barriers to entry. It-has also been pointed out (namely in Pindyck [1991] and Dixit [1992b]) that this same theoretical framework can be applied to the cases of financial investments and bank credit, to the extent that the presence of a "lemon premium" may cause financial investments to be partially irreversible. In many cases, the stochastic behaviour of the relevant state variable describing the "on-going" uncertainty (in our case xt) has been described by a (Bertola [1991], geometric Brownian motion and Caballero Dixit [1992a], Pindyck [1988]). The implications of these results for our analysis are quite straightforward. To the -extent that the future value of the bank investment assets is uncertain, the decision is partially irreversible because it carries some kind of sunk costs. Therefore, it seems reasonable to assume that for some financial investments 12 See, for example Bertola and Caballero [1991] 114 the "value of waiting" is greater than the value of "having in hand" an investment project (i. e. a financial asset in our case). In such a context, the bank needs a sort of "buffer stock" asset; the investment into a "buffer stock" asset could be assimilated to a "non-investment" decision, taken in order to "wait" for more information about those specific financial investments that contain sunk costs. Chamley [1993] provides a rigourous formalization of the investment decision in a model where the investment in liquid assets corresponds to "non-investment". In his model, postponing investments enables the agents to gather information and to avoid investments in bad states. Chamley proves (among other results) that °... if the time interval to reverse the no-investment position is sufficiently short, the immediate investment by all agents cannot be an equilibrium outcome, because in this outcome, each agent has an incentive to postpone his investment in order to learn first the impact of aggregate investment on his individual payoff" (Chamley [1993] ). Assuming that the commercial banks' free liquid reserves play the role of temporary liquid assets, their demand will be higher the higher the "value of waiting", i. e. the higher the degree of uncertainty about the future value of financial investments. Since a big portion of banks' assets is constituted by direct it credit to enterprises, can be argued that the aggregate level of banks' free liquidity ratio (which is an aggregate stock of temporary liquid assets held by the banking system) might be affected by all those variables affecting the general degree of risk of the economy. An approximate measure of the risk of the economy 115 might be given by the rate of variation of the price level, since it typically carries uncertain and asymmetric effects on the economy as a whole. In such a context, if the yield on the free liquid reserves is given by the "value of waiting" for additional information on the financial investments, then the rate of growth of the price level should act as a proxy for the yield on the free liquid reserves. As a consequence, the free liquidity ratio should be positively correlated with the rate of growth of the price level. Obviously, the size of a temporary asset (like the free liquid reserves) might be affected by the volume of transactions performed by the banks and by the volume of credit intermediated. -A variable that might capture the effect of such a "transactionary" motivation on the free liquidity ratio might be the gross margin of banks' intermediation, or a proxy for it. In the empirical specification that follows, it has been assumed that the free liquid reserves of commercial banks behave at the same time as a financial market "buffer stock" asset (i. e. affected by considerations on the profitability of the financial investments and by the "value of waiting" for further information on the degree of riskiness of financial investments) and as a "transactionary" liquid asset, whose size is affected by the volume of transactions, determined by the margin of intermediation. 4. The model The free liquidity ratio is defined as the ratio between the free reserves held by the commercial banks at the central banks and the total deposits. The estimates are based on quarterly data and 116 have been implemented by following the "general-to-specific" methodology. The definitions of all the variables considered and all the tables with all the estimates and diagnostic tests are shown in the Appendix. One of the main difficulties in building our model lies in the relevant institutional modifications that affected the Italian context throughout the sample period considered. In particular, among the most relevant phenomena, we can quote the gradual increase of importance of the stock market, and the birth of new kinds of financial intermediaries. Like in Richter and Teigen, it will be assumed that the free liquidity ratio for commercial banks depends on the level of the required reserves ratio, on the variations of the liquidity stock determined by the State and by the foreign sector, on the liquid opportunity cost of holding reserves, and on the yield of reserves (or on the convenience of holding them). The opportunity cost is typically described by a representative money market interest rate. For what concerns the own yield of liquid reserves, while Richter and Teigen consider that the free liquid reserves are higher most required the the variance of the reserve requirement, in this work it will be assumed - on the basis of the above-mentioned literature on investment under uncertainty and with sunk costs - that the free liquidity ratio depends positively on the degree of uncertainty of the whole economy, summarized by the-rate of growth level. On hand, of the price the other as far as a transactionary liquid is motivation for the assets also present, the free liquidity be ratio should positively correlated with the margin of intermediation - acting as a proxy for the volume of banks' 117 financial intermediation - given that the higher the profitability higher incentives to increase of the bank -intermediation, the the the existing stock of financial funds intermediated by the banks. If the financial investment of the banks is regarded as partially irreversible and subject to uncertainty, and is assumed to be alternative to the choice of keeping the funds invested in liquid liquidity be assets, then the free ratio should positively "value " for further information correlated to the of waiting about the new, financial investments, as argued above. If the "value of to the degree in the waiting" is a concept associated of uncertainty the higher the degree the economy (in the sense that of uncertainty, higher the value of "waiting for further information"), then the free liquidity ratio should be positively correlated with the rate level, be here for of growth of the price which will used as a proxy the degree of uncertainty of the whole economy. The° variable is the the implicit employed for this purpose rate of variation of deflator of the gross domestic product, henceforth DINF. In fact, it is the rate of growth of the price level, rather than the price level itself, that best captures the concept of "on-going increase level (i. uncertainty", since a constant on the price e. a in the be non-stationary trend prices time series) could associated to a situation of constant inflation and static expectations, while, hand, it is is in on the other reasonable to expect that it a trend the Increase in the price levels (i. e. a trend in the rate of growth index) that directly of the price more and explicitly affects the degree of uncertainty for the agents of the economy', by making more difficult for them to make inferences on the continuously changeable the different causal links existing among variables. Since the rate 118 of growth of the price level is calculated with respect to two consequent periods of time, a qualitatively similar information could be captured by a suitable lag structure of the price level. Therefore, an alternative specification has also been implemented, which includes the price level (whose suitable lag structure will again be determined by following the general-to-specific approach) instead of its rate of growth DINF. In the present model, the dependent variable is defined, like in Richter and Teigen [1982], as the ratio between the free bank banks' reserves - or unconstrained reserves - and total bank deposits, i. e. the current account, saving deposits, plus the deposit certificates of central banks. In the estimates, such a W. variable is defined as Like Richter and Teigen, we assume that free reserves are affected by the behaviour of reserves requirements, but, unlike Richter and Teigen, the variance of reserves requirements is not regarded here as a proxy for the yield (or convenience) of holding free reserves, but, like in Richter and Teigen, it is still assumed that higher levels of reserve requirements would reduce the need for free reserves as a liquid liquidity asset, given that would in that case be less scarce (since the banks might be able, in case of need, to temporarily and costly diverge from the reserve requirements). For this reason, it is free liquidity assumed here that the ratio is negatively correlated with the coefficient of reserves requirements (defined as L1). Using the levels of L1 instead of its variance as a regressor constitutes, in our opinion, a more general approach: in fact, since we are following the general-to specific methodology, and our general four unrestricted model contains lags, if it is the variability 119 level this rather than the of reserve requirements that matters, final dynamic the should be evident from the structure of model, if the which could suggest, necessary, a re-specification of whole diagnostic it model, according to the results of the tests. However, is assumed here that the required reserves are relevant also because the higher their level, the more costly (in terms of opportunity liquid because in cost) it is to keep free reserves, this way the banks have to waive a larger amount of potentially profitable financial investments. For this purpose, the variable considered here is defined as L1, and is the ratio between the constrained deposits. As be later, reserves and the total will pointed out and lagged L1 also in the tables of the appendix, the values of are not four lags. This significant in the general unrestricted model with be in confirms the fact that the variable Li seems to significant its level rather than in its variance. Like in Richter and Teigen, it is assumed that the free liquidity ratio is affected by the variations in the aggregate liquidity determined by the public sector and the foreign sector. the two variables here considered for this purpose are quite similar to the ones employed by Richter and Teigen: the first is defined as between G and is the ratio the variation of the credits of the Bank Public Sector bank of Italy with the and the average level of deposits over the considered period; the second, defined as FX, is the ratio between the variation of the net foreign position of the bank of Italy and the average level of deposits over the considered period. As mentioned earlier, and on the basis of the literature on investments irreversibility, we assume that the free liquidity ratio 120 higher degree should be higher the the of uncertainty and risk in characterizing the whole economy. Therefore, since the present "value is (as discussed model the proxy for the of waiting" above) DINF, then the dependent variable L3 is expected to be positively correlated with DINF. However, an alternative specification of a lag general unrestricted model containing a polynomial in the deflator) variable P (the implicit GDP instead of DINF will be considered. The results of such an alternative specification are - as we will see later - analogous to those of the specification with DINF. Finally, the variable RDIFF3, defined as the difference between the rate of interest on the bonds issued by special credit institutions (RBCI) and the rate of discount (RD), has been considered. There are several reasons that justify a variable so defined. First of all, RBCI can be regarded as a representative typical medium-long term interest rate, while RD is - obviously -a money market interest rate. Therefore RDIFF3 contains information on the time structure of the interest rate, whose modifications may determine re-allocations of the banks' assets. To the extent that the free reserves are a temporary liquid asset, their relative size should be positively correlated to the volume of funds re- by banks. allocations performed the Secondly, RDIFF3 could be regarded as a proxy for banks' because, being determined profitability by the difference between a long run and a short run interest rate (and containing therefore information on the term structure of the interest rates), it provides information about one of the most typical roles of the banking system: transforming the maturity and degree of risk of 121 financial assets. In addition, in Italy, for a large part of the time series here considered, many small banks operating only locally had been investing most of their funds in assets issued by special credit institutions operating in the medium and long run. This situation gave rise to the phenomenon known in the Italian context as "double intermediation". To the extent that RDIFF3 contains information on banks' profitability, it should be regarded as an opportunity cost of holding free reserves, and should be expected to be negatively correlated with W. On the other hand, if the free liquid reserves are mainly a temporary liquid asset, determined by the process of funds re-allocation, we should expect the variations of RDIFF3 to be because positively correlated with L3, changes in the term structure of the interest rate should determine a re-allocations of the banks' financial funds. The signs of the various coefficients of the lag polynomial of RDIFF3 can tell us whether free liquid reserves have to be interpreted as a temporary asset held in the process of transactions and re-allocations, or according to the wealth allocation theory, which would suggest that RDIFF3 plays the role of an opportunity be cost, and should therefore negatively correlated with W. The time series considered for the estimates start in 1975, in order to exclude the year of the first oil shock, 1974. Since have been quarterly data employed for the estimates, a general four lags unrestricted model with has been defined, since, as argued by the econometric literature, such a number of lags is sufficient to "capture" the dynamic behaviour of a model defined over quarterly data. 122 In some preliminary analyses, a lag polynomial of the three months treasury bill, which is one of the most relevant money market interest rates in the Italian context (and which had been successfully employed in a previous version of the present model, estimated over a shorter sample period), turned out to be non significant in a general model including lag polynomials in L1, FX, G, RDIFF3, or in L1, FX, G, RDIFF3, or in Li, FX, G, DINF. Other variables (as the treasury bills rate, the call money rate, the interbank rate, the interest rate on long term government bonds, the stock market index and its variance over the last twelve months, or its variation weighted with the variance over the last twelve months) that have been regarded as' proxies for the opportunity costs in other empirical works on banks' liquid reserves turned out to be non significant or less significant. The package MICROFIT version 3.0 has been employed for the estimates. The appendix contains the definition of all the variables and the tables containing the results of the estimates and the tests in detail. Following the general-to-specific methodology, the following general unrestricted model with four lags has been estimated. 444 L3 = const +E ai L3t-i +E Ni FXt-i +E Ti Gt-i + i=1 i=0 1=0 44 +E ni RDIFF3t-i +E 6i DINFt-i + et; with e- N(o, a2) (1 i=0 i=0 The data employed for the estimates start from 1975, first quarter, but, since the model employed for the estimates contains four lags, the period covered by the estimates starts in 1976, first 123 quarter. The estimates have been run over the sample period 1976 QI- 1991 QIV, and over the sample period 1976 QI - 1989 QIV, both for the general unrestricted model and for the restricted one. The latter sample period has been considered in order to be able to perform the predictive failure test, although this implies a loss of degrees of freedom. Table 1 in the appendix contains the results of the estimates and diagnostic tests of the general unrestricted model over the sample period 1976 QI - 1989 QIV. In the Chi-square version of the functional form test, HO is rejected both at the level of confidence of 0.95 and 0.99; in the F-version of the same test, HO is marginally rejected at the level of confidence of 0.95, while it is not rejected at the level of confidence of 0.99. The failure of the functional form test, which usually indicates that the function is not properly specified, might be due, in this case, to the fact that the general unrestricted model has a low number of degrees of freedom over the sample period 1976 QI - 1989 QIV, since the results in this regard are much better in the restricted model. and over the sample period 1976 QI - 1991 QIV, both for the general unrestricted and for the restricted model. The general unrestricted model estimated over the sample period 1976 QI - 1989 QIV also marginally fails, at the level of confidence of 0.95, both the Chi-square and the F version of the heteroscedasticity test, while HO is rejected at the level of confidence of 0.99. This, again, might be caused by the low number of the degrees of freedom, since the general unrestricted model, when estimated over the complete sample period, largely passes the 124 Heteroscedasticity tests both in the Chi-square and in the F 0.95. version, at the level of confidence of However, Table 1 shows that the remaining diagnostic tests (serial correlation, normality, predictive failure) do not present any problem over the sample period 1976 QI -1989 QIV. Table 2 shows the variable deletion test on the restrictions on the general model that determines the final "parsimonious" specification, shown in detail in table 3 and briefly reported here. L3 = 0.019335 -0.085682 L1 +0.067034 FX +0.099713 G+ +0.0010731 RDIFF3 +0.10260 DINF-3 -0.0015241 RDIFF3-4 + +0.25831 L3-4 (2 R2 = 0.65962 Table 3 shows that the specification suggested in the the restricted model largely passes all the diagnostic tests at level of confidence of 0.95. A first look at the dynamic specification obtained suggests that the behaviour of the free liquidity ratio is inconsistent with the assumption (made also by Richter and Teigen [1982] a priori) of instantaneous adjustment of the free liquid reserves. Such an by the hypothesis assumption was motivated of rational expectation in financial markets. This does not mean that the rational hypothesis expectation must necessarily be rejected, but it suggests that some delays in the adjustment of the variables (consistent with the some form of rationality of agents) exist. . quation 2 can be reparametrized according to the following pattern: 125 L3 = const + ßp Li + ß1 FX + ß2 G+ ß3 RDIFF3 + +ß4 DINF-4 + ß5 (RDIFF3-RDIFF3-4) + ß6 L3-4 (3 Equation 4 yields the following equation, estimated over the sample period 1976 QI - 1989 QIV: L3 = 0.019335 -0.085682 L1 +0.067034 FX +0.099713 G+ -0.0004511 RDIFF3 +0.10260 DINF_3 +0.25831 L3-4 + 0.0015241 (RDIFF3-RDIFF3_4); (4 According to the theoretical background presented in the previous (and partly in the present) section, the coefficient p3 of the variable RDIFF3 is expected to be negative, since it may be regarded as an opportunity cost for holding liquid assets. On the other hand, the coefficient ß5 of the term (RDIFF3-RDIFF3-4) is expected to be positive, since it could be associated to a variation in the potential bank profitability, or in the term structure of the interest rate. In fact, such phenomena would determine a re- financial increase allocation of the bank's assets, which could the amount of liquidity necessary for the bank's transactionary purpose. The lagged value of L3 appearing in all of the restricted be determined specifications may by adjustment costs. The sign of the coefficient referred to DINF is positive. This increase in suggests that the the price level may not be regarded as but an opportunity cost, can rather be interpreted in accordance to the literature on investments with sunk costs and under uncertainty. In fact, if the bank credit and some kinds of financial assets are irreversible regarded as (partially) investments, the higher the the level of uncertainty and risk of whole economy - signaled by the level rate of growth of the price DINF - and the higher the "value 126 of waiting", the higher. the incentives for the banks to keep some funds in perfectly liquid temporary (and costlessly reversible) assets, such as the free liquid reserves held at the central bank. Intuitively, one could think of a "transactionary" component of the demand for banks' liquid reserves, positively correlated with the term (RDIFF3-RDIFF3-4), and a "speculative" component positively correlated with DINF. On the other hand, the present level of the variable RDIFF3 could be regarded as an opportunity cost for holding reserves. Table 4 shows the general unrestricted model estimated over the complete sample period 1976 QI - 1991 QIV, which yields satisfactory results in almost all of the diagnostic tests, except the Chi-square version of the functional form test, at the level of confidence of 0.95 (while for the same test H0 is rejected at the level of confidence of 0.99), and the f version of the same test, at the level of confidence of 0.95. Table 5 contains the variable deletion test which determines the final specification, and shows that the general unrestricted model estimated over the sample period 1976 QI - 1991 QIV supports the same restrictions as the one estimated over the sample period 6 1976 QI - 1989 QIV. Table shows the final specification of the model, the same as the one obtained for the sample period 1976 QI - 1989 QIV, which yields satisfactory results in all of the diagnostic tests. The values of the parameters of the restricted model seem to be very close in the estimates performed in the two different sample periods. Even in the estimates using the whole sample (i. e. 1976 QI 127 the be to - 1991 QIV), restricted model can reparametrized according the pattern of equation 3, obtaining the following estimate. L3 = 0.023203 -0.114752 L1 +0.076871 FX +0.078367 G+ -0.0008477 RDIFF3 +0.10131 DINF_3 +0.30852 L3_4 + 0.0016986 (RDIFF3-RDIFF3_4); (5 R2 = 0.68225 Since the variable DINF is defined as the rate of variation of the price level, a similar kind of information - as far as the dynamic behaviour of the price level is concerned - should be obtained by introducing in an ADL model a lag polynomial in the price level P, instead of a lag polynomial in the variable DINF. Table 7 shows a model of this kind estimated over the sample period 1976 QI - 1989 QIV, and table 10 over the sample period 1976 QI - 1991 QIV. In both cases the general unrestricted model yields satisfactory results for almost all of the diagnostic tests, excepting the functional form test. Table 8 (for the sample period 1976 QI - 1989 QIV) and table 11 (for the sample period 1976 QI 1991 QIV) show the variable deletion tests performed in order to test a restricted model containing P-3 and P_4 instead of the three-period-lagged rate of (DINF_3). variation of the prices (for Table 9 the sample period 1976 QI - 1989 QIV) and table 12 (for the sample period 1976 QI - 1991 QIV) show the final specification containing the two lagged variables P-3 and P-4 instead of DINF-3. As is shown in both tables, the restricted model passes almost all of the diagnostic tests, excepting (marginally) form test, the the functional over sample period 1976 QI - 1989 QIV. The better performances of the model containing the variable DINF 128 could suggest that this variable better captures the "on-going uncertainty", associated to the "value of waiting" that could be imagined as the original motivation for the liquidity preference of the bank. However, apart from the functional form diagnostic test, the results of table 9 are very similar to those of table 3, and those of table 12 are very similar to those of table 6, as expected. This suggests that the specification obtained from a model including a lag polynomial in the variable DINF is analogous and consistent with the specification obtained from a general unrestricted model with a lag polynomial in the price level. 5. Conclusions. The empirical results presented here suggest that the behaviour of the banks' free liquidity ratio can be interpreted according to a theoretical framework based on the implications of the recent literature on investments under uncertainty and with sunk costs. justifies Such a framework - for the banks -a particular concept of "speculative" and "transactionary" demand for money, which shows many similarities with the Keynesian liquidity preference theory. In fact, in this context, the free liquid reserves could be interpreted as temporary and liquid assets that the banks hold in the process of collecting information about (partially). irreversible financial investments, or in the process of reallocating their funds. According to the theoretical framework followed in this paper, the amount of free liquid reserves depends on the "value of waiting" for new information about the possible investments in (less liquid) in financial assets. Increases the "value of waiting" may be due to 129 the increase in the degree of uncertainty and riskiness of the whole economy such as perceived by the banks. In this model it has been assumed that the rate of growth of the price level (the variable DINF) or, alternatively, some measure of variation of the price level - obtained form the general-to-specific methodology can be regarded as a proxy for the degree of uncertainty and risk of the aggregate economy. The "value of waiting" may also be correlated with the increase in the profitability of the banks (summarized in our specific case by the difference RDIFF3-RDIFF3-4, ), to the extent that this might generate a process of reallocation of the financial investments of the banks, requiring a temporary "transactionary" liquid asset. In our specific case, the difference RDIFF3-RDIFF3-4, such as determined by the general-to-specific methodology, may be regarded as the variation of the banks' potential profitability, since the variable RDIFF3 is a proxy for the banks' gross margin of intermediation. The theoretical approach to the banks' behaviour followed in this chapter and supported by the empirical analysis carries some relevant macroeconomic implications. In particular, the results are consistent with the "credit view" and with those macroeconomic models (like Bernanke and Blinder [1988], described in the chapter 1 of this thesis) which emphasize the willingness of banks to invest and their perception of the degree of risk and uncertainty of the whole economy, and attribute to these factors a key role in the determination of the aggregate level of credit, liquidity and output. 130 Bibliography of Chapter 4 Bera, A. K., Jarque, C. M. [1981], "An Efficient Large-Sample Test for Normality of Observations and Regression Residuals, Working Paper in Econometrics no. 40, Canberra, Australian National University. Bernanke, B. S. [1983], "Non-Monetary Effects of the Financial Crisis in the Propagation of the Great Depression", American Economic Review, vol. 73, pp. 257-276. Bernanke, B. S. [1990], "On the Predictive Power of Interest Rates NBER Working Paper 3486, Cambridge MA. Spread" , no. Bernanke, B. S., Blinder, A. S. [1988], "Is It Money, Credit or Both or Neither? Credit, Money and Aggregate Demand", American Economic Review, vol. 78, pp. 435-451. Bernanke, B. S., Blinder, A. S., [1990], "The Federal Funds Rate and the Channels of Monetary Transmission" NBER Working Paper no. 3487, Cambridge, Ma. Bertola, G., Caballero, R. J., [1991], "Irreversibility and Aggregate Investment", NBER Working Paper no. 3865. Blinder, A. S., [1987], "Credit Rationing and Effective Supply Failure", Economic Journal, vol. 97, pp. 327-352. Blinder, A. S., Stiglitz, J. E., [1983], "Money, Credit Constraint and Economic Activity", American Economic Review, vol. 73, pp. 297- 302. Bryant, R. C., [1983], "Controlling Money: The Federal Reserve and Its Critics", Washington D. C., Brookings Institutions. Cagan, P. [1969], "Interest Rates and Bank Reserves -A Reinterpretation of the Statistical Association", in Essays on Interest Rates, ed. by J. M. Guttentag and P. Cagan, New York, NBER, pp. 223-272. Chamley, C. [1993], "Liquidity Trap", mimeo, presented at the Warwick Summer Research Workshop "The New Macroeconomics: Imperfect Markets and Policy Effectiveness", University of Warwick, 12-30 July 1993. Cuthbertson, K. [1985], "Banking Lending to U. K. Industrial and Commercial Companies", Oxford Bulletin of Economics and Statistics, vol. 47, pp. 91-118. Deshmukh, S. Greenbaum, S. I., Kanatas, G. [1983], "Bank Forward Lending in Alternative Funding Environments", Journal of Finance, vol. 38, pp. 873-886. Diamond, P. A., [1984], "Financial Intermediation and Delegated Monitoring", Review of Economic Studies, vol. 51, pp. 393-414. 131 Dixit, A., [1989], "Entry and Exit Decision Under Uncertainty", Journal of Political Economy, vol. 97, pp. 620-638. Dixit, A., [1992a], "Investment and Hysteresis", Journal of Economic Perspectives, vol. 6, pp. 107-132. Dixit, A., [1992b], "Irreversible Investment With Uncertainty and Scale Economies", mimeo, Princeton University, Duca, J. V., Vanhoose, D. D., [1990], "Loans Commitments and Optimal Monetary Policy", Journal of Money, Credit and Banking, vol. 22, pp. 178-194. Dufour, J. M., [1980], Dummy Variables and Predictive Tests for Structural Changes, Economic Letters, vol. 6, pp. 241-247. Gertler, L. G., [1988], "Financial Structure and Aggregate Economic Activity: an Overview", Journal of Money, Credit and Banking, vol. 20, pp. 559-588. Godfrey, L. G., [1978a], "Testing Against General Autoregressive and Moving Average Error Models When the Regressors Include Lagged Dependent Variables", Econometrica, vol. 46, pp. 1293-1301. Godfrey, L. G., [1978b], "Testing for Higher Order Serial Correlation in Regression Equations When the Regressors Include Lagged Dependent Variables", Econometrica, vol. 46, pp. 1303-1310. Gurley, J., Shaw, E., [1955], "Financial Aspects of Economic Development", American Economic Review, vol. 45, pp. 515-538. Jarque, C. M., Bera, A. K., [1980], "Efficient Tests for Normality, Homoscedasticity and Serial Independence of Regression Residuals", Economic Letters, vol. 6, pp. 255-259. Kaldor, N., [1982], "The Scourge of Monetarism", Oxford, Oxford University Press. Kindleberger, C., [1978], "Manias, Panic and Crashes", New York, Basic Books. Langhor, H., [1981], "Banks Borrowing From Central Bank and Reserve Position Doctrine: Belgium 1960-1973". Journal of Monetary Economics, vol. 7 pp. 107-124. Minsky, H. [1975], "John Maynard Keynes", New York, Columbia University Press. Pindyck, R., [1988], "Irreversible Investment, Capacity Choice and the Value of the Firm", American Economic Review, vol. 78, pp. 969-985. R., [1991], "Irreversibility, Uncertainty Investment", -Pindyck, and Journal of Economic Literature, vol. 29, pp. 1110-1148. Poitervin, M., [1989a], "Collusion and the Banking Structure of a Duopoly", Canadian Journal of Economics, vol. 22, pp. 263-277. 132 Poitervin, M., [1989b], "Financial Signaling and the Deep-Pocket Argument" Rand Journal of Economics, vol. 20, pp. 26-40. Poitervin, M., [1990], "Strategic Financial Signaling", International Journal of industrial organization, vol. 8, pp. 499-518. Ramsey, J. B., [1969], "Tests for Specification Errors in Classical Linear Least Squares Analysis", Journal of the Royal Statistical Society, series B, vol. 31, pp. 350-371. Ramsey, J. B., [1970], "Models, Specification Error and Inference: a Discussion of Some Problems in Econometric Methodology" Bulletin of the Oxford Institute of Economics and Statistics, vol. 32, pp. 301-318. Richter, R., Teigen, R. L., [1982], "Commercial Bank Behaviour and Monetary Policy in an Open Economy: West Germany 1960-1980", Journal of Monetary Economics, vol. 10, pp. 383-405. Salkever, D. S., [1976], The Use of Dummy Variables to Compute Prediction Errors and Confidence Intervals", Journal of Econometrics, vol. 4, pp. 393-400. Vanhoose, D. D., [1985], "Bank Market Structure and Monetary Control", Journal of Money, Credit and Banking, vol. 17, pp. 298-311. Wessels, R. E., [1982], "The Supply and Use of Central Bank Advance Facilites", Journal of Monetary Economics, vol. 10, pp. 89-100. 133 APPENDIX List and description of the variables BLIQ = Commercial Banks free reserves held at the Bank of Italy (also defined as BLIQ = BRES - REQRES); BRES = BLIQ + REQRES; DBITE = FBITES-FBITES-1; DEP = Current account, saving deposits and certificates of deposit of commercial banks; DEPAV = (DEP + DEP_1)/2; DFX = FXBI - FXBI-1; DINF = (P - P-1)/P-1; DBITE = Credits of the Italian Central Bank with the Italian Treasury; G= DBITE/DEPAV; FXBI = net foreign position of the Italian Central Bank; L1 = REQRES/DEP; L3 = BLIQ/DEP; P= implicit deflator of the Italian GDP; RBCI = rate of interest on the bonds of special credit institutions; discount; RD = official rate of RDIFF3 = RBCI-RD; REQRES = required reserves of the commercial banks; The source of the data is "Supplemento al Bollettino Statistico della Banca d'Italia", with the exception of the variables P and RBCI, whose time series have been provided by DATASTREAM Services (at the Univewrsity of Warwick), on the basis of OECD and Bank of Italy data. Diagnostic Tests The diagnostic test performed in the tables that follows are those provided by the package MICROFIT 3.0, briefly described in the appendix to the third chapter. TABLE 1 Dependent Variable : L3 (ordinary least squares) 56 Observations : 76Q1 - 89Q4 Regressor Coefficient Std. Error t-ratio CONST 0037698 025605 -. . -. 14723 L1 31016 17114 -. . -1.8123 FX 064257 041937 1.5322 . . G 18029 059156 3.0478 . . RDIFF3 0010357 6947E-3 1.4909 . . DINF 12189 083007 -. . -1.4684 L3-: 086127 19064 45178 L -. . -. L1-i 094389 10124 93320 . . . FX-i 023941 038019 62970 . . . G-3. 092621 060976 1.5190 . . RDIFF3-i 1067E-3 7204E-3 14815 . . . DINF-i 10944 082970 -. . -1.3190 L3-2 083835 15496 54100 . . . L1-2 015078 10908 13823 . . . FX-2 062467 040045 1.5599 . . G-2 014034 062966 22289 . . . RDIFF3-2 7158E-3 7266E-3 98512 . . . DINF-2 13966 079568 1.7553 . . L3-3 22209 16504 1.3457 . . L1-3 062563 10775 58065 -. . -. FX-3 057762 038407 -. . -1.5039 G-3 069076 053503 -. . -1.2911 RDIFF3-3 9141E-3 7106E-3 -. . -1.2864 DINF-3 26125 080241 3.2559 . . L3-4 39815 15193 2.6206 . . L1-4 28406 18816 1.5096 . . FX-4 047411 038738 -. . -1.2239 G-4 066266 052619 -. . -1.2594 RDIFF3-4 0015988 7358E-3 -. . -2.1729 DINF-4 010158 072631 13986 -. . -. 80569 R-squared . F- test(29,26) 3.7175 Residual 2902E-3 Regression S. E. 0033409 sum of sq. . . S. D. dep. var.. 0052110 Mean dep. 013338 of . of var. . DW - stat. 1.7622 Max. log-likelihood 261.3076 Diagnostic tests Test statistics LM version F verpion Serial correlation Chi-sq( 4) 8.3230 F( 4,22) 96014 . Functional form Chi-sq( 1) 8.2662 F( 1,25) 4.3294 Normality Chi-sq( 2) 2.7169 not applicable Heteroscedasticity Chi-sq( 1) 4.3516 F( 1,54) 4.5498 Predictive failure Chi-sq( 8) 8.3876 F( 8,26) 1.0484 TABLE 2: Variable Deletion Test (OLS) Dependent Variable : L3 (ordinary least squares) List of the variables to be deleted from the regression: DINF L3-i L1-i FX-1 G-i RDIFF3-i DINF-i L3-2 L1-2 FX-2 G-2 RDIFF3-2 DINF-2 L3-3 L1-3 FX-3 G-s RDIFF3-3 L1-4 FX-4 G-4 DINF-4'- 56 Observations used for estimation from 76Q1 to 89Q4 Regressor Coefficient Std. Error t-ratio CONST 019335 011803 1.6381 . . L1 085682 061803 -. . -1.3864 FX 067034 030737 2.1809 . . G 099713 035553 2.8046 . . RDIFF3 0010731 3833E-3 2.7999 . . 10260 051390 1.9966 "DINF-3 . . RDIFF3-4 0015241 3806E-3' -. . -4.0045 L3-4 25831 094211 2.7418 . . Joint test of zero restrictions on the coefficients of the deleted variables: Lagrange Multiplier Statistic CHI-SQ(22): 24.0317 Likelihood Ratio Statistic CHI-SQ(22): 31.3940 F Statistic F(22,26) 88842 : . Table 3 Dependent Variable : L3 (ordinary least squares) 56 Observations : 76Q1 - 89Q4 Regressor Coefficient Std. Error t-ratio CONST 019335 011803 1.6381 . . Li 085682 061803 -. . -1.3864 FX 067034 030737 2.1809 . . G 099713 035553 2.8046 . . RDIFF3 0010731 3833E-3 2.7999 . . DINF-3 10260 051390 . . 1.9966 RDIFF3-4 0015241 3806E-3 -. . -4.0045 L3-4 25831 094211 2.7418 . . 65962 F 7,48) R-squared . - test( 13.2887 Residual 5084E-3 Regression S. E. 0032543 sum of sq. . . S. D. dep. var.. 0052110 Mean dep. 013338 of . of var. . DW - stat. 2.0719 Max. log-likelihood 245.6106 Diagnostic tests Test statistics LM version F version Serial correlation Chi-sq( 4) 5.9674 F( 4,44) 1.3120 Functional form Chi-sq( 1) 2.7788 F( 1,47) 2.4540 Normality Chi-sq( 2) 2.4170 not appl icable Heteroscedasticity Chi-sq( 1) 3.1793 F( 1,54) 3.2502 Predictive failure Chi-sq( 8) 7.6287 F( 8,48) 95359 . TABLE 4 Dependent Variable : L3 (ordinary least squares) 64 Observations : 76Q1 - 91Q4 Regressor Coefficient Std. Error t-ratio 015878 019723 80504 CONST . . . 20250 13861 L1 -. . -1.4608 089845 039545 2.2720 FX . . 12392 053117 2.3330 G . . 4820E-3 6231E-3 77354 RDIFF3 . . . 085172 074139 DINF -. . -1.1488 015069 16242 092777 L3-1 -. . -. 026848 091035 29492 L1-i . . . 0063814 036042 17705 FX-1 . . . 034838 050171 69439 G-1 . . . 3810E-4 6737E-3 056555 RDIFF3-i -. . -. 046302 074278 62337 DINF-i -. . -. 10135 13861 73119 L3-2 . . . 022660 089866 25209 L1-2 -. . -. 037804 036066 1.0482 FX-2 . . 014769 051956 28427 G-2 -. . -. 2243E-3 6777E-3 33101 RDIFF3-2 . . . 11549 072098 1.6019 DINF-2 . . 32542 14593 2.2299 L3-3 . . 035279 086196 40929 L1-3 -. . -. 051524 034605 FX-s -. . -1.4689 093654 046633 G-3 -. . -2.0083 9606E-3 6983E-3 RDIFF3-3 -. . -1.3756 19516 069629 2.8028 DINF-3 . . 37860 13897 2.7243 L3-4 . . 12062 14188 85014 L1-4 . . . 040350 036806 FX-4 -. . -1.0963 090393 046476 G-4 -. . -1.9449 0019906 6784E-3 RDIFF3-4 -. . -2.9342 014627 064359 22711 DINF-4 -. . -. 79299 F- R-sciuared . test(29,34) 4.4912 3838E-3 Regression Residual sum of sq. . S. E. 0033599 0054250 Mean . S. D. of dep. var.. . of dep. var. 012533 1.8960 Max. . DW - stat. log-likelihood 293.9631 Diagnostic tests Test statistics LM version F version Chi-sq( 4) 5.4641 F( 4,30) 70009 Serial correlation . Functional form Chi-sq( 1) 5.8889 F( 1,33) 3.3442 Normality Chi-sq( 2) 5.1442 no t applicable Heteroscedasticity Chi-sq( 1) 2.2943 F( 1,62) 2.3053 TABLE 5: Variable Deletion Test (OLS) Dependent Variable : L3 (ordinary least squares) List of the variables to be deleted from the regression: DINF L3-i Li-z FX-i G-i RDIFF3-i DINF-i L3-2 L1-2 FX-2 G-2 RDIFF3-2 DINF-2 L3-s L1-3 FX-3 G-3 RDIFF3-3 L1-4 FX-4 G-4 DINF-4 64 Observations used for estimation from 76Q1 to 91Q4 Regressor Coefficient Std. Error t-ratio 023203 0098608 2.3530 CONST . . 11475 Li -. _050595 -2.2680 076871 029799 2.5797 FX . . 078367 031487 2.4888 G . . 8509E-3 3498E-3 2.4326 RDIFF3 . . 10131' 048108.2.1060 r DINF-3 . . 0016986 3699E-3 RDIFF3-4 -. . -4.5919 30852 086850 3.5523 L3-4 . . Joint test of zero restrictions on the coefficients of the deleted variables: Lagrange Multiplier Statistic CHI-SQ(22): 22.3052 Likelihood Ratio Statistic CHI-SQ(22): 27.4245 F Statistic F(22,34) 82676 : . f Table 6 Dependent Variable L3 (ordinary least squares) 64 Observations : 76Q1 - 91Q4 Regressor Coefficient Std. Error t-ratio CONST 023203 0098608. 2.3530 . . L1 11475 050595 -. . -2.2680 FX 076871 029799 2.5797 . . G 078367 031487 2.4888 . . RDIFF3 8509E-3 3498E-3 2.4326 . . DINF-3 10131 048108 2.1060 . . RDIFF3-4 0016986 3699E-3 -. . -4.5919 L3-4 30852 086850 3.5523 . . 68225 R-squared . F - test(29,34) 17.1771 Residual 5891E-3 Regression S. E. 0032435 sum of sq. . . S. D. dep. var.. 0054250 Mean dep. 012533 of . of var. . DW - stat. 1.9467 'Max. log-likelihood 280.2508 Diagnostic tests Test statistics LM version F version Serial correlation Chi-sq( 4) 6.5191 F( 4,52) 1.4744 Functional form Chi-sq( 1) 46637 F( 1,55) 40373 . . Normality Chi-sq( 2) 5.1661 not appl icable Heteroscedasticity Chi-sq( 1) 2.9345 F( 1,62) 2.9794 TABLE 7 Dependent Variable : L3 (ordinary least squares) 56 Observations : 76Q1 - 89Q4 Regressor Coefficient Std. Error t-ratio CONST 088081 054782 -. . -1.6078 L1 019680 21159 093008 -. . -. FX 045630 046148 98879 . . . G 18519 060375 3.0673 . . RDIFF3 0010022 7014E-3 1.4288 . . p 5119E-3 0012622 40556 -. . -. 071497 18747 38138 L3-i . . . 32218 15231 2.1153 Li-i . . 029063 038270 75943 FX-i . . . 080636 061162 1.3184 G-i . . 1226E-3 7592E-3 16149 RDIFF3-i . . . 2297E-3 0016815 13662 P-1 . . . 15190 16786 90493 L3-2 . . . 13172 14056 93712 L1-2 . . . 050455 039503 1.2772 FX-2 . . 013772 57196 24079 G-2 -. . -. 4703E-3 7348E-3 64007 RDIFF3-2 . . . P-2 0030587 0018026 1.6969 . . 19322 17076 1.1315 L3-3 . . 085022 12049 70566 L1-3 . . . FX-3 059549 037958 -. . -1.5688 G-. 082098 052031 s -. . -1.5779 RDIFF3-3 0013316 7432E-3 -. . -1.7917 0015483 0017813 86924 P-3 . . . L3-4 32562 16212 2.0085 . . L1-4 17166 18877 90938 . . . FX-a 060466 039796 . -. . -1.5194 G-4 -. 045978 050575 -. 90910 RDIFF3-4 0012928 . 7808E-3 -. . -1.6558 P-4 0047610 0015283 -. . -3.1153 79099 R-squared . F - test(29,26) 3.3929 Residual of 3122E-3 Regression S. E. 0034650 sum sq. . . S. D. dep. var.. 0052110 Mean dep. 013338 of . of var. . DW - stat. 1.8073 Max. log-likelihood 259.2647 Diagnostic tests Test statistics LM version F versio n Serial correlation Chi-sq( 4) 4.4100 F( 4,22) 47015 . Functional form Chi-sq( 1) 15.7970 F( 1,25) 9.8233 Normality Chi-sq( 2) 80449 icable . not appl Heteroscedasticity Chi-sq( 1) 2.0559 F( 1,54) 2.0580 Predictive failure Chi-sq( 8) 8.4959 F( 5,26) 1.0620 TABLE 8: Variable Deletion Test (OLS) Dependent Variable : L3 (ordinary least squares) List of the variables to be deleted from the regression: P L3-i L1-i FX-i G-s RDIFF3-i P-1 L3-2 Ll-2 FX-2 G-2 RDIFF3-2 P-2 L3-3 L1-3 FX-3 G-3 RDIFF3-3 L1-4 FX-4 G-4 56 Observations used for estimation from 76Q1 to 89Q4 Regressor Coefficient Std. Error t-ratio 022851 011802 1.9362 CONST . . 069551 078939 88107 L1 -. . -. 076169 031505 2.4177 FX . . G 092573 035773 2.5878 . . RDIFF3 97577E-3 4035E-3 2.4179 . . P-3 0010826 7893E-3 1.3715 . . 0011457 8115E-3 P-4 -. . -1.4119 RDIFF3-4 0016866 4163E-3 -. . -4.0513 L3-4 22610 093781 2.4109 . . Joint test of zero restrictions on the coefficients of the deleted variables: Lagrange Multiplier Statistic CHI-SQ(21): 22.4504 Likelihood Ratio St atistic CHI-SQ(21): 28.6902 F Statistic F(21,26) 82849 : . .17 Table 9 Dependent Variable : L3 (ordinary least squares) 56 Observations : 76Q1 - 89Q4 Regressor Coefficient Std. Error t-ratio CONST 022851 011802- 1.9362 . . Li 069551 078939 88107 -. . -. FX 076169 031505 2.4177 . . G 092573 035773 2.5878 . . RDIFF3 97577E-3 4035E=3 2.4179 . . P-3 0010826 7893E-3 1.3715 . . P-4 0011457 8115E-3 -. . -1.4119 RDIFF3-4 0016866 4163E-3 -. . -4.0513 22610 093781 2.4109 L3-4 . . 65112 F 7,48) R-squared . - test( 10.9646 Residual of 5211E-3" Re S. E. 0033296 sum sq. . gression . S. D. dep. 0052110 Me dep. 013338 of var.. . an of var. . DW - stat. 2.0608 Ma x. log-likelihood 244.9196 Diagnostic tests Test statistics LM version F version Serial correlation Chi-sq( 4) 5.2847 F( 4,43) 1.1202 Functional form Chi-sq( 1) 5.9604 F( 1,46) 5.4793 Normality Chi-sq( 2) 1.4179 not appl icable Heteroscedasticity Chi-sq( 1) 2.3787 F( 1,54) 2.3955 Predictive failure Chi-sq( 8) 4.0846 F( 8,47) 51058 . TABLE 10 Dependent Variable : L3 (ordinary least squares) 64 Observations : 76Q1 - 91Q4 Regressor Coefficient Std. Error t-ratio 027609 035760 77208 CONST -. . -. 13364 14692 909618 LI -. . -. 062601 044485 90961 FX . . -. 11718 052976 2.2120 G . . 7645E-3 6298E-3 1.2139 RDIFF3 . . 8587E-3 9488E-3 90501 p -. . -. 0013699 16293 0084079 L3-i . . . 11704 11454 1.0218 Li-i . . 0084395 036263 23273 FX-i . . . 039619 053049 74685 G-1 . . . 9685E-3 7091E-3 13658 RDIFF3-i . . . 0011425 0013807 82745 p-1 . . . 071539 13782 51908 L3-2 . . . 048689 10574 46045 L1-2 . . . 027483 036561 75171 FX-2 . . . 0038531 050140 076846 G-2 -. . -. 1221E-3 6928E-3 17625 RDIFF3-2 . . . 0014074 0014722 95597 P-2 . . . 22319 14767 1.5114 L3-3 . . 088914 10468 84938 L1-3 . . . 044169 035229 FX-3 -. . -1.2538 075577 045169 G-3 -. . -1.6732 0012158 7317E-3 RDIFF3-3 -. . -1.6615 9361E-3 0014937 62666 P-3 . . . 31198 14413 2.1645 L3-4 . . 16089 15120 1.0641 L1-4 . . 052233 038418 FX-4 -. . -1.3596 G-4 -. 054721 045103 -1.2132 0014132 . 7362E-3 RDIFF3-4 -. . -1.9197 0028906 0011715 P-4 -. . -2.4673 77662 4.0762 R-squared . F - test(29,34) 4142E-3 Regression S. E. 0034902 Residual sum of sq. . . 0054250 Mean 012533 S. D. of dep. var.. . of dep. var. 1.8245 . DW - stat. Max. log-likelihood 291.5277 Diagnostic tests Test statistics LM version F versi on '30) Serial Chi-sq( 4) 5.6936 F( 4, 73237 correlation . Functional form Chi-sq( 1) 11.4200 F( 1,33 ) 7.1674 Normality Chi-sq( 2) 1.6863 not applicable Heteroscedasticity Chi-sq( 1) 1.7314 F( 1,62) 1.7239 TABLE 11: Variable Deletion Test (OLS) Dependent Variable : L3 (ordinary least squares) List of the variables to be deleted from the regression: P L3-i L1-i FX-i G-i RDIFF3-i P-i L3-2 L1-2 FX-2 G-2 RDIFF3-2 P-2 L3-3 L1-3 TX-3 G-3 RDIFF3-3 L1-4 FX-4 G-4 64 Observations used for estimation from 76Q1 to 91Q4 Regressor Coefficient Std. Error t-ratio CONST 020968 0095354 2.1989 . . Li 050117 064032 78269 -. . -. FX 081942 029312 2.7955 . . 083338 031282 2.6641 G . . RDIFF3 83547E-3 6880E-3 1.3681 . . 94138E-3 6880E-3 1.3681 P-3 . . P-4 0010241 7027E-3 -. . -1.4574 0018415 3734E-3 RDIFF3-4 -. . -4.9314 L3-4 25501 084727 3.0097 . . Joint test of zero restrictions on the coefficients of the deleted variables: Lagrange Multiplier Statistic CHI-SQ(21): 17.1962 Likelihood Ratio Statistic CHI-SQ(22): 20.0268 Statistic F(22,26) 59485 F : . Table 12 Dependent Variable : L3 (ordinary least squares) 64 Observations : 76Q1 - 91Q4 Regressor Coefficient Std. Error t-ratio CONST 020968 0095354 2.1989 . . 050117 064032" _ L1 -. . -. 78269 FX 081942 029312 2.7955 . . G 083338 031282 2.6641 . . RDIFF3 83547E-3 6880E-3 1.3681 . . P-3 94138E-3 6880E-3 1.3681 . . p-4 0010241 7027E-3 -. . -1.4574 RDIFF3-4 0018415 3734E-3 -. . -4.9314 L3-4 25501 084727 3.0097 . . 69455 F 8,55) R-squared . - test( 15.6330 Residual of 5663E-3 Re S. E. 0032089 sum sq. . gression . S. D. dep. 0054250- Me dep. 012533 of var.. . an of var. . DW - stat. 2.0484 Max. log-likelihood 281.5144 Diagnostic tests Test statistics LM version F version Serial correlation Chi-sq( 4) 5.2806 F( 4,51) 1.1466 Functional form Chi-sq( 1) 2.84264 F( 1,54) 2.5099 Normality Chi-sq( 2) 3.1569 not appl icable Heteroscedasticity Chi-sq( 1) 3.0690 F( 1,62) 3.1229 OBS. BLIQ DEP FBITES FXBI P RBCI 75Q1 2459.0 82485.0 21642.9 391.2000 22.5000 10.4800 75Q2 2636.0 85676.0 23063.6 441.4000 22.9000 10.7800 75Q3 3861.0 90621.0 24995.9 -633.3000 23.4000 10.8000 75Q4 3414.0 99854.0 29533.0 -1366.6 24.0000 10.6600 76Q1 1993.0 103828.0 32561.7 -3161.4 25.0000 12.6000 76Q2 1865.0 105436.0 36003.6 -3318.8 26.7000 13.8700 76Q3 1398.0 109025.0 37168.2 -2726.4 28.0000 13.7100 76Q4 2383.0 121569.0 39511.0 5205.9 29.6000 14.1800 77Q1 2444.0 125453.0 38733.0 5683.6 --31.2000 14.8800 77Q2 1683.0 129355.0 35035.4 7870.5 31.9000 14.6000 77Q3 2346.0 133667.0 34436.0 9986.1 33.1000 14.4400 77Q4 3236.0 149886.0 35217.0 11899.6 34.2000 13.9800 78Q1 4238.0 154717.0 38354.1 13242.5 35.4000 13.6000 78Q2 2389.0 159981.0 35599.8 15707.6 36.3000 13.2300 78Q3 1972.0 165802.0 34251.1 18952.5 37.8000 12.9300 78Q4 5059.0 185027.0 40026.0 19659.9 38.9000 13.3900 79Q1 2339.0 185895.0 35316.7 24811.8 40.2000 13.4400 79Q2 1793.0 191252.0 33883.2 28465.5 42.0000 13.4200 79Q3 1699.0 197757.0 34245.6 28465.5 43.7000 13.4100 79Q4 3875.0 221879.0 40692.0 30538.0 46.0000 14.2700 80Q1 2582.0 216952.0 41554.6 36932.0 47.8000 14.7600 80Q2 2823.0 219232.0 44460.9 45771.5 50.3000 15.2000 80Q3 2971.0 221608.0 46678.9 52004.6 52.4000 15.9600 80Q4 4958.0 251264.0 50262.0 55068.1 54.6000 16.3000 81Q1 3681.0 245411.0 56604.3 54548.1 57.0000 17.6700 81Q2 4057.0 242977.0 58947.3 56968.4 59.3000 20.9500 81Q3 5061.0 242296.0 61510.7 55597.4 62.3000 21.3300 81Q4 6150.0 274127.0 64179.0 58058.4 65.4000 21.0000 82Q1 3358.0 266007.0 66406.0 50926.9 67.8000 20.4300 82Q2 5410.0 270194.0 67349.0 48003.7 69.9000 20.7500 82Q3 4785.0 279505.0 67024.0 50371.0 73.0000 19.8000 82Q4 4455.0 325080.0 75751.0 51166.0 75.8000 19.8600 83Q1 4834.0 313167.0 82012.0 58064.0 78.6000 18.2800 83Q2 5679.0 318257.0 75787.0 67666.0 80.8000 17.9000 83Q3 2490.0 329890.0 75280.0 71889.0 83.4000 17.4700 83Q4 3931.0 368389.0 80181.0 76089.0 86.4000 17.3302 84Q1 2851.0 354292.0 85252.0 71999.0 89.1000 15.250e 84Q2 3637.0 354612.0 81368.0 74815.0 91.0000 14.7302 84Q3 2293.0 365653.0 82467.0 79801.0 92.7000 14.450e 84Q4 5008.0 411862.0 90101.0 81813.0 94.6000 13.840E 85Q1 2302.0 404952.0 100235.0 76337.0 97.4000 12.780¬ 85Q2 3594.0 407547.0 99824.0 82401.0 99.2000 13.420E 85Q3 2790.0 418497.0 103329.0 60960.0 101.0000 13.080¬ 85Q4 7968.0 454170.0 117563.0 61435.0 102.3000 13.270¬ 86Q1 4040.0 436713.0 128705.0 66826.0 104.7000 12.550E 86Q2 4433.0 437883.0 117894.0 66260.0 107.4000 9.800E 86Q3 3662.0 447918.0 121062.0 60960.0 109.1000 9.540E 86Q4 4699.0 496101.0 128503.0 61435.0 110.2000 9.0504 87Q1 4782.0 482944.0 130153.0 66826.0 111.2000 9.130¬ 87Q2 3727.0 490784.0 132631.0 65871.0 114.0000 9.920E 87Q3 4527.0 494098.0 141973.0 65318.0 115.0000 11.020E 87Q4 5647.0 531819.0 137223.0 74305.0 117.0000 11.190¬ 88Q1 5582.0 509205.0 137680.0 74829.0 118.7000 10.8404 88Q2 4932.0 519164.0 137192.0 74128.0 120.7000 10.850( OBS. BLIQ DEP FBITES FXBI P RBCI 88Q3 6151.0 531417.0 140600.0 78782.0 123.4000 11.0800 88Q4 4899.0 571564.0 139446.0 82823.0 124.9000 11.0500 89Q1 7829.0 555000.0 139103.0 88890.0 126.5000 11.5700 89Q2 4239.0 565720.0 131363.0 91306.0 128.4000 11.7700 89Q3 4458.0 570320.0 131321.0 98938.0 130.2000 11.6700 89Q4 5453.0 625348.0 145469.0 92875.0 132.7000 12.0800 90Q1 8159.0 598021.0 145104.0 98003.0 135.8000 12.1500 90Q2 4072.0 607436.0 124856.0 110079.0 137.8000 12.0700 90Q3 4105.0 619686.0 130429.0 109929.0 139.8000 12.0500 90Q4 4471.0 686279.0 146388.0 103335.0 143.3000 11.2900 91Q1 5227.0 649292.0 148311.0 111080.0 145.2000 12.1600 91Q2 1854.0 662288.0 137574.0 109322.0 148.0000 11.2400 91Q3 2926.0 670060.0 144041.0 108950.0 151.2000 11.5100 91Q4 4849.0 746907.0 163898.0 95342.0 152.9000 11.4300 92Q1 *NONE* *NONE* *NONE* *NONE* *NONE* 11.1400 . _ý OBS. RD REQRES 75Q1 8.0000 11159.0 75Q2 7.0000 11508.0 75Q3 6.0000 12032.0 75Q4 6.0000 12040.0 76Q1 12.0000 14382.0 76Q2 12.0000 14755.0 76Q3 12.0000 15167.0 76Q4 15.0000 16622.0 77Q1 15.0000' 18383.0 77Q2 13.0000 19081.0 77Q3 11.5000 19571.0 77Q4 11.5000 20262.0 78Q1 11.5000 22774.0 78Q2 11.5000 23678.0 78Q3 10.5000 24389.0 78Q4 10.5000 25281.0 79Q1 10.5000 27816.0 79Q2 10.5000 28653.0 79Q3 10.5000 29513.0 79Q4 15.0000 30671.0 80Q1 15.0000 32527.0 80Q2 15.0000 32434.0 80Q3 16.5000 33009.0 80Q4 16.5000 33749.0 81Q1 19.0000 36924.0 81Q2 19.0000 36113.0 81Q3 19. 35844.0 81Q4 19.0000"0000 36654.0 82Q1 19.0000 40982.0 82Q2 19.0000 40379.0 82Q3 18.0000 42226.0 82Q4 18.0000 45184.0 83Q1 18.0000 50979.0 83Q2 17.0000 49955.0 83Q3 17.0000 52905.0 83Q4 17.0000 54108.0 84Q1 16.0000 60042.0 84Q2 15.5000 58555.0 84Q3 16.5000 60306.0 84Q4 16.5000 62838.0 85Q1 15.5000 71656.0 85Q2 15.5000 70726.0 85Q3 15.5000 73125.0 85Q4 15.0000 74890.0 86QI 14.0000 79583.0 86Q2 12.0000 78582.0 86Q3 12.0000 79200.0 86Q4 12.0000 83858.0 87Q1 11.5000 89626.0 87Q2 11.5000 90740.0 87Q3 12.0000 91107.0 87Q4 12.0000 92438.0 88Q1 12.0000 96198.0 88Q2 ý 12.0000 96201.0 0135. RD REQRES 88Q3 12.5000 99516.0 88Q4 12.5000 101822.0 89Q1 13.5000 106250.0 89Q2 13.5000 107413.0 89Q3 13.5000 110452.0 89Q4 13.5000 111370.0 90Q1 13.5000 117756.0 90Q2 12.5000 116793.0 90Q3 12.5000 119399.0 90Q4 12.5000 124322.0 91Q1 12.5000 128756.0 91Q2 11.5000 122010.0 91Q3 11.5000 121804.0 91Q4 12.0000 127762.0 92Q1 12.0000 *NONE* 134 CHAPTER 5 OPTIMAL PHYSICAL CAPITAL AND FINANCIAL STRUCTURE OF THE FIRM: TWO FACES OF THE SAME DECISION? *I am very grateful to Keith Cowling and Norman Ireland for their helpful suggestions. I am also indebted to Giovanni Amisano, Flavio Rovida and Jeremy Smith for their comments and suggestions regarding the econometric part of this work, and to Pier Luigi Sacco, Jeffrey Bernstein and Colin Mayer for helpful comments and/or discussions at various stages of my work. Obviously, none of the above-mentioned people are responsible for any mistakes that might be found or for the views expressed here. Many thanks also to Dr. Giovanni Bucchieri and to the "Ufficio Studi" of Mediobanca for providing me with the volumes of the data employed for the empirical analysis of this chapter. 135 OPTIMAL PHYSICAL CAPITAL AND FINANCIAL STRUCTURE OF THE FIRM: TWO FACES OF THE SAME DECISION? 1. Introduction. The present chapter and the next one are closely related, since they constitute two different parts of the same investigation. The analysis has been divided into two parts in order to simplify the exposition and in the attempt of making it less tedious. The introductory comments contained in this section refer therefore both to this chapter and chapter 6. The purpose of the analysis is to provide a simplified dynamic framework in which the decisions concerning the firm's investment and financial structure take place simultaneously, in the presence of bankruptcy and adjustment costs and with imperfect competition in the goods market. The assumption of imperfect competition has been introduced in order to capture a few strategic effects of the investment decision, although with a few simplifying assumptions. However there are a few relevant differences between the approach followed in this chapter and in chapter 6. ' The theoretical part of this chapter uses - as a tool of interpretation -a model by Bernstein and Nadiri [1986], with a few modifications, mainly consisting in the introduction of imperfect competition in the goods market, while some more substantial modifications have been introduced in the empirical implementation of the model. In the model by Bernstein and Nadiri [1986] the cost of is financial capital a function of the leverage ratio and- the firm's decisions concerning the financial structure and investments i35. a take place at the same time, but the variables affecting the leverage ratio are assumed to be exogenous. In this way the profits do not interact with the leverage ratio and with the cost of financial capital. In the model of chapter 6, the firm chooses simultaneously the level of investments as well as the optimal financial structure, and the cost of financial capital still depends on the financial structure of the firm, but, in this case, the cost of capital, the optimal financial structure and the profits are assumed to interact, because of a few possible causal links suggested by various theoretical contributions (from finance and industrial economics) briefly analyzed in chapter 6. If, under a given set of assumptions (that will be discussed in this chapter and in the next one), a link between the profits and the leverage ratio of the firm exists, then that the by degree - to the extent profits are affected the of concentration and other strategic variables - the simultaneous decision problem of investment and optimal financial structure is also affected by those market strategic interactions that imperfect characterize a context of competition. While this last point is one of the main concerns of chapter 6, the present chapter will follow the theoretical part of the contribution by Bernstein and Nadiri [1986], as a simplified starting point for the empirical firms data. analysis based on The empirical analysis includes a few estimates of an investment function based on three industrial samples of data of Italian firms operating respectively in the chemical, electronic, and clothing sectors. The source of the data is the survey provided 13S$ by Mediobanca (an Italian credit institution), a reliable and commonly employed source of firm's data. The relevance of the firm's financial structure for the investment decision, with asymmetric information on financial markets, has been the subject of many contributions both in finance and in industrial economics, although most of them do not raise the problem of simultaneity between investment and financial structure. 136 In izlustrial economics, the relevance of firms' financial structure has been pointed out - within the context of predatory pricing models - by the literature concerned with the "deep pocket argument" (Telser [1966], Benoit [1984], Poitervin [1989a]) and by the models based on the assumption of "limited liability effect" (Brander and Lewis [1986], Poitervin [1989b] ). In the former, it is assumed that "strong" firms can afford a long-during price war because they can rely on large financial resources. In the latter, a high level of debt is regarded as a pre-commitment for an aggressive policy. The signalling game of entry deterrence yields - as a result financial for the incumbent the - the optimal structure and entrant. Following the famous contribution by Fazzari, Hubbard and Petersen [1988] that shows the significance of cash flow for the firm investment decision, the relevance of the firm financial by structure has been analyzed in many neoclassical studies introducing some form of market imperfection and/or transaction costs within a Tobin's Q framework. The focus was often on R&D investments (Bernstein and Nadiri [1986] and [1993]), or, in another class of investment models, on a hierarchy of finance funds, resulting from transaction costs in placing new shares and taxes (see for example Bond and Meghir [1992]). A causal and recursive link between financial structure and investments decision had already been pointed out in the past by the "radical economics" literature, and, more recently, by some "new- Keynesian" contributions in a macroeconomic framework (like for instance Greenwald and Stiglitz [1988], [1989], [1990a], [1990b], Bernanke and Gertler [1989]). 137 Regarding the internally generated flow of profits as the main source of finance, as a result of the imperfect substitutability between internal and external funds for the firm, the radical economists often envisage a causal link between the flow of profits for the firm and the investment decision. Finally, a link between financial structure and investment decision may simply be based on an incentive argument which can be summarized by considering that with asymmetric information "... the greater the debt-equity ratio, the more the incentives of managers who act in the interest of equity holders diverge from the interests of creditors" (Fazzari, Hubbard and Petersen, [1988], p. 151). This last argument has led to many other contributions, even based on different theoretical approaches, to formalize the cost of financial capital (and, as a consequence the rate of discount of future incomes for the representative firm) as an increasing function of the leverage ratio or the gearing ratio (for example Bernstein and Nadiri [1986], Ganoulis [1991], Bond and Meghir [1992]), or to include managerial costs depending on the amount of borrowing in the profit function (Bernstein and Nadiri [1993]). The recent new-Keynesian contributions aim at formalizing the information role of asymmetric and risk in affecting the willingness of financial intermediaries to supply credit and its cost. Most of the empirical analyses on firm investment decision that are based on some version of Tobin' sQ theory rely on stock market data and assume perfect competition on the goods market. A model that might be potentially adapted to describe an imperfectly competitive framework is the one by Abel and Eberly [1993], which introduces a stochastic variable in the firm's profits, and is meant 138 to provide a framework that incorporates fixed cost investments in a stochastic context with investment irreversibility. Another purpose of this paper is to attempt to overcome a limitation of the empirical works based on Tobin's Q theory, which rely on small samples mainly composed by firms issuing shares on the stock market: such firms usually enjoy a well established reputation, and may not always be regarded as representative for the entire population. In Italy (like in all of the non-securitized financial systems) the firms operating in the stock market only constitute a very small minority. In order to use the data from the Mediobanca sample (mainly accounting data) some restrictions will have to be made. However, for what concerns the general issue of the use of accounting data, this paper follows Martin [1993], who, objecting to Benston's [1982] criticism on the reliability of accounting data, argues as follows: "... If this argument were correct, the consequences would be severe indeed. It would mean that industrial economists could not carry out empirical research. It would mean that a wide spectrum of government publications describing economic activity ought to be discontinued, since they are based on what is, originally, accounting data. " (Martin, 1993, p. 517). The next section (and its subsections) contains the first model, mainly based on Bernstein and Nadiri [1986]. Section 3 contains the comments on the empirical analysis, section 4 contains the conclusions. A few drawbacks and limitations of Bernstein and Nadiri's framework are illustrated in the next chapter, which contains a brief digression on a few issues concerning the financial decisions of the firm. 139 2. A model of financing and investment with asymmetric information and bankruptcy costs This model resembles, in its theoretical part, the one by Bernstein and Nadiri [1986], with a few modifications mainly consisting in two aspects: the first aspect is the above-mentioned introduction of imperfect competition in the goods market. The second is the introduction of a distinction between "purely technological" adjustment costs of investment, and "strategic" costs of investments. Such a distinction is not very relevant for the theoretical part of the analysis (since a few simplifying assumptions are made on the conjectures of the different firms), but is going to carry some significant consequences for the empirical analysis that will be implemented in this paper. The model provides a framework to formalize the interaction between financial structure and investment decisions. Following Bernstein and Nadiri [1986], it is assumed that financial and real decisions are connected in the sense that in each moment in time the firm chooses the financial structure that minimizes the cost of is capital. It the minimized cost of capital that affects in its turn the investment decision of the firm. As in Bernstein and Nadiri [1986], the decision of each firm is assumed to take place in but, in continuous time, unlike Bernstein and Nadirs, it is revealed to other firms and outsiders only at discrete intervals. This last assumption will carry some relevant implications for the empirical later. analysis, as we will see Again as opposed to Bernstein and Nadirs [1986], the firms act in a context of imperfect competition. 140 Let us define the variable profits as: var. profits of firm i= u°(k, wlpi) (1 where k= physical capital; w= variable inputs cost; determined by demand pi = profit margin, the elasticity, product differentiation, conjectures, and other aspects affecting the market power of the firms. It is assumed that in each moment the firm is optimizing the inputs is quantity' of variable employed, whose price w. The mark up depend firms pi is assumed to on the conjectures of 'the and on the demand elasticity. As a first approximation, it is assumed that the different firms is that conjectures of are symmetrical. It assumed increasing levels of k always increase the variable profits, i. e. the variable profits depend positively on the capital stock. This is means that the marginal profitability of capital positive even is when the average profitability of capital negative. Let us also assume that the following budget constraint holds: k(t) = S(t) + B(t) (2 the S the B the debt. where K is physical capital, value of shares, Broadly speaking, this is equivalent to saying that the short run the liabilities, that assets equal short run or the payments are implemented instantaneously. The firm accumulates capital according to the following rule: k=I- gk, k(0)>0 and lim k(t)? (3 t->oD -O depreciation I is where g is the rate of of capital stock, the investment level. The rate of depreciation is assumed to be fixed, that 05g51. A dot for and it is also assumed over a variable stands 141 differentiation with respect to time. The flow of funds, which must hold at any moment in time, is defined as follows: 0(0(t))]. u°(k, wtpi) - A[I(t)] - [rf(t) + B(t) + B(t) + (4 + S(t) - D(t) =0 where u°(k, wipi) is the, same as defined earlier, A[I(t)] is the function of (purely technological) adjustment costs of investments, D(t) the dividends, rf(t) the interest rate on risk free borrowing, ß(t)=B(t)/S(t) the leverage ratio, $(R(t)) a risk premium on borrowing, which is a monotonically increasing function of the leverage ratio, B(t) is the level of borrowing, S(t) the value of the new shares issues. A[I(t)] is twice continuously differentiable, A">O, to the with A(0)=0, A'>O, according standard assumptions. A[I(t)] is also assumed to be increasing, monotonic and invertible. As in Bernstein and Nadirs, the function $(n(t)) represents the bondholders, depends leverage premium paid to which on the ratio, O(0) is differentiable 0(0)=0, such that twice continuously and 0'>0. The function of the risk premium O(Q) recognises the'fact that the cost, of debt increases as the firm increases its leverage. Bernstein and Nadiri's assumption that the firm operates in the interests of its shareholders is kept (in contrast with the model of the next chapter). Therefore, like in Bernstein and Nadiri, in each their flow period, the shareholders equate rate of return to the of funds accruing to them. Formalizing: rs = D(t)/S(t) + S(t)/S(t) (5 is the where rs rate of return on shares and S is the value of the that the includes shares. This means rate of return on shares the 142 dividends per share plus the capital gain. Bernstein and Nadiri define then V=B+S as the value of the financial capital of the firm. Departing from Bernstein and Nadiri, in this paper the value of shares S is not defined on the basis of stock market data, since in the sample employed for the empirical analysis we are dealing mainly with firms not issuing shares on the stock markets, and operating in a country (Italy) where the stock market is of very little relevance, since the magnitude of the transactions concerning securities is very small compared to the magnitude of intermediated financial funds. However this distinction is only relevant for what concerns the empirical analysis, and will be explained later in more detail. Using equations 4,5, and the definition V=B+S, following Bernstein and Nadiri [1986], we get: V= [(rs + rf"C + O(n)-n) / (1 + ()]V- r (6 where r= u°(k, wlpi) - A(I) (7 6 Equation shows that the financial capital value of the firm due may increase to the cost of capital and the variable profits net of "purely technological" adjustment costs of investments. Bernstein and Nadiri, by integrating equation 6 obtain the following general solution: ttt Jr(r)exp [_J(s)ds]dt} V(t) = expI P(-c)dt] {c- (8 0 "c" is the where constant of integration and M)/(l+n)] is the =['(rs+rfO+O(O cost of financial capital. In order to have 143 lim V(t)>O with D>O t->4)0 Bernstein and Nadiri impose that 00 t Jr(t)exP c [_Jsds]dt (9 00 Therefore, at time t=0, equation 8 becomes the following 00 Z JexPEJ(s)ds]. V(O) = [r(t)dt] (10 00 The firm, in order to maximize. the value of the initial financial capital, minimizes 4 by selecting the optimal financial V(O) by level structure, and maximizes choosing the optimal of investments, given the optimal variable profits. The right-hand side 10 is functional. of equation the objective In this framework, the first, the is procedure is recursive: unit cost of financial capital by the financial minimized, second, using minimized unit cost of initial capital, the present value of the variable profits, net of the "purely technological" adjustment costs of investments, is "financial' decision is described follows: maximized. The as O(n)"o) -t° = min [(rs + rf"n + / (1 + n)] ill Q hence: d-P O'A(1+A) = [-rs + rf + + 0] / (1+A)2 =0 (12 do In this framework the value of the shares and its instantaneous variation are exogenous, as well as the interest rate on the risk free debt rf, while the risk premium 0 depends on the leverage ratio n. the optimal financial structure corresponds to a situation where 144 the marginal rate of return to shareholders equates the marginal cost of debt, which is determined by the interest rate on risk free assets and the marginal premium required by lenders. By substituting equation 12 into equation 11, we get: -,i, ° = rf + q'n + $M (13 The minimized cost of financial capital (D° such as defined in equation 13, is the rate of discount of variable profits, net of "purely technological" adjustment costs. Therefore, the problem of determining the optimal level of investments may be redefined as follows: co z JexP(S)dS]. Eu0(t)-A(I)]dt J° = max (14 I 00 subject to conditions 3. The optimal conditions can be obtained from the following Hamiltonian: H= u°(t) - A(I(t)) + q(I-gk) (15 Assuming that the second order conditions are satisfied, the first order conditions are the following: 8H/8I = -A' +q=0 (16 q= (''° + 9)q- 8u°/Sk (17 where q is the shadow price of the constraint, defined by using the current marginal valuations. Conditions 16 and 17, must hold 3. together with conditions The transversality conditions will be the following: lim z(t) ? 0, [k*(t)-k(t)]z(t) =0 (18 145 Assuming where z(t) - e-4°tq(t). that the transversality conditions 16 17 be by the are satisfied, equations and can solved obtaining following: A"I = (4°+g)A' - 6u°/dk (19 Condition 19 corresponds exactly to the standard results of the firm's investment decision, neoclassical models on with the only difference that here the rate of discount 4° depends on the leverage in its depends few ratio, which, turn, on a exogenous variables information the financial summarizing the concerning markets. Graphically, the equilibrium may be represented as follows: FIGURE 1 I > K In the graphic, SS is the stable saddlepath. It is important to in this is note that 6u°/Sk case not the marginal productivity of 146 capital of the standard neoclassical perfectly competitive models of investment decision, but it is the marginal profitability of capital, which is conditional on the (exogenous) conjectures. Therefore,, in this- framework, disturbances in the optimal level of investments may be caused not only by exogenous technological by in shocks, but also changes the conjectures, market shares, demand elasticity, as well as in the (exogenous) "financial" variables, like the value of shares and the interest rates on risk free assets. 2.1 The empirical implementation Figure 1 shows that in the long run equilibrium E both conditions 3 and 19 are satisfied. From equation 19, we get: Aý _ (6u°/8k)/(, D°+9) hence: I* = G[(6u°/8k)/(4°+g)] (20 level where I* is the optimal of investments, and G= A'-l. The function G is monotonic and increasing, since the same has been is assumed about A(I(t)). I* conditional on the exogenous mark up (which, in its turn, might depend on the demand elasticity, and interest market power) as well as the rate on risk free assets and the (exogenous) rate of return and value of shares. We assume that for each firm the decisions are taken in continuous time, but made known to outsiders at discrete intervals. In other words, each firm its investment reveals to the others (and financial structure) decision only when the balance sheet and profits and losses account Nevertheless, the investment are published. decision is taken in 147 continuous time, although the information on the rivals' behaviour is founded on the latest available accounting data, which constitute the information source for conjectures. Following Cowling's [1982] interpret approach, one can the values of the firms' conjectures as an interval whose lower bound is given by the total absence of higher bound collusion, and the by the situation of collusion. We assume therefore that the conjectures of the firm appearing in 4 equations 3 and are consistent with a game admitting a sequential equilibrium, corresponding to a situation of collusion. If a firm deviates from its observable levels of investments of the previous is be periods (which assumed to a source of information for the conjectures of the other firms), it might cause the other firms to revise their strategy and, as a consequence, their conjectures. A deviation from the path, of investments might induce the rival to change their behaviour, and possibly follow a less collusive and it more aggressive policy. Therefore, might be possible to think of investments' a particular category of adjustment costs - defined here as "strategic" adjustment costs of investments' for convenience deriving from the fact that firm - as those costs when a modifies its level of investments, it might have to bear the costs determined by changes in the rivals' strategy. In analogy with other empirical (mainly the specifications ones with partial adjustment, containing the lagged dependent variable among the regressors) we can think of the "costs of being, out of the optimal level of investment" as "strategic opposed to the costs of adjustment", that, in this framework, would be proportional to the deviation of the investments from level, at time "t" the previous at time "t-1". 148 At this point we need a few approximations and simplifying assumptions in order to perform the empirical analysis. First of all, we will indicate analytical forms for the function A(I) (describing the "purely technological" adjustment costs of investments) and formalize the functional link between the marginal profitability of capital and the average profitability of capital. Both formalizations need to be consistent with all of the assumptions of the model. What we need is simply a monotonically increasing one-to-one function between the marginal profitability of capital and the average profitability of capital. For what concerns the "purely technological" adjustment cost of investments we need a function with the following characteristics: A(0)=0, A'>O, A">0. A fairly general approximation for the function A(I), could be the following, which also has the advantage of being easy to handle on the algebraic point of view: A= 00"Ia with a>l; 00>0 (21. For what concerns the functional link between 6u/6k and u/k, it has to be consistent with the assumption, previously expressed, that the marginal profitability of capital is always positive, even when the average profitability of capital is negative. This is because the marginal profitability of capital is not empirically observable, unlike the average profitability of capital. Therefore, a condition for the latter to be employed as an approximation for the former is the existence of a one-to-one invertible functional link between them. The following analytical form shows the properties of having a positive marginal profitability of capital - even when the average is profitability of capital negative - and, at the same time, of 149 showing an invertible one-to-one functional link between the average profitability of capital and the marginal profitability of capital: 8u/8k = exp(al"u/k) (22. Substituting definitions 21 and 22 in equation 20, taking logs and rearranging yields the following equation: 1 al u1 In I* = ln(1/e0"a) +- in(-, D°+g) (23 a- 1 a- 1k a- 1 The variable P° is not observable and equation 13 expresses V as a function of the exogenous rate of return on shares and the leverage ratio A. Given that g is assumed to be constant and fixed, the leverage ratio n could be used as a proxy for the two variables in the bracket on the right-hand side of equation 23. In particular, we will use 61"0 as a proxy for the variable (4'°+g). Substituting then gl. QP for (4)°+g) in equation 23, we obtain the following: 1 al uR In I* _ [ln(1/eoa)+lne1] +- "ln(n) (24 a-1 a-1 k a-1 Where ß depends on the relation between 61"nß and (, °+g). The size ß is of the parameter not stated a priori, and will be determined by the data. In this way, the data themselves will specify whether the (4°+g) is i. variable defined as concave or convex in ei. Such a formulation allows us not to rule out a priori neither a risk averse lover the lenders. nor a risk attitude of We need now to take into account the "non-technological" investments. adjustment costs of It is assumed that, as in the model the firm of partial adjustment, minimizes a quadratic cost function including the cost of being out of equilibrium as well as the deviating from strategic cost of the level of investment signalled 150 for formalized at the previous time. The problem the firm may be as follows: min C= a(It - I*)2 + b(It - It-1)2 (25 C is the total "non-technological" cost. The variables written in italic (the investments) are expressed in logs, "a" represents the "cost of being out of equilibrium", "b" represents the "non technological cost of deviating from the previous level of investments". Assuming that the second order conditions for the described by 25 optimization problem equation are satisfied, we obtain the following first order conditions: öC/oIt = 2a(It - I*) + 2b(It - It-1) =0 (26 It = [a/(a+b)]"I* + [b/(a+b)]"It-1 =u I* + (1 - p)It-1 (27 27 where p=a/(a+b). Substituting equation into equation 24, we get: u In It = ß0 + PI - R2"ln(R) + (33 In It-1 (28 k where N U al Po= Lln(1/60a)+ln911; Pi= :" ß2=(ß"u)/(a-1); ß3=1-µ a-1a-1 Equation 28 represents the specification employed for the econometric estimations. The value of the coefficients of equation 29 might be interpreted in economic terms. Looking at equations 27 be that, and 28, it could said at time "t", the higher the "non- technological" cost of deviating from the "t-1" level of investments, the lower the value of the parameter p of equation 27, in which must be, any case, greater than 0 and smaller than 1. To in industry give an example, an characterized by a high degree of the firms, collusion among and an extremely high expected cost of 151 breaking the collusion, the parameter p should be positive and close to 0. In the extreme case, it should be equal to 0. This would imply that the coefficient ß3 of the lagged investments in equation 28 should be equal to 1, and equation 28 could be reformulated in the following way U. aln It = ß0 + ßl - R2"ln(n) + (28' . k but the regressors should be (in this extreme case) non significant. A negative value, or a value greater than one for the coefficient ß3, would in any case contradict the theoretical framework of this paper. No particular restrictions can be made on the value of the (u/k) coefficient of and ln(n), apart from the fact that the former has to be positive, the latter negative, and their absolute value larger (i. should be smaller, the ß3 e. the closest to 1 is ß3). 2.2 The dataset and the use of the data The dataset has been constructed by using all of the available firms' data from the Mediobanca sample, for three industries: the chemical, the electronics and the clothing. The appendix contains a followed report on the methodology in processing the data and the list of all of the firms included in the sample. A few comments concerning the definition of B(t) and k(t) are necessary at this point. Most empirical work on investments and firms' financial structure employ data obtained from the stock market and tend not to use accounting data. Often, in those cases, the samples include only a small number of firms, usually the ones issuing securities in the stock and bond markets. It must be said, 152 however, that these firms represent a small minority, both in terms of their number and their volume of trade, in a country with a non- securitized financial system. Since one of the purposes of this paper was to perform estimates on a very large and representative sample of firms operating in a few given industries, a different be followed. approach had to The sample provided by Mediobanca is one of the most complete sources of information. It was therefore a natural choice for the present purpose. For the sake of the empirical analysis we have to consider that institutional our data refer to an context (Italy) where the larger overwhelmingly amount of credit is provided by the banking interest system, usually at a variable rate. Therefore, to the interest is flexible extent that the rate and adjusts, the balance be sheet value could a reasonable approximation for the "market value" of debt, although some distortion might derive from the for absence of a measure the risk of insolvency. In addition, if banks are regarded as institutions obtaining economies of scale in collecting information and monitoring the performances of the borrowers, then the information they have about the quality of their customers is not only private, but it is also part of the the banks' entrepreneurial skills of managers. To the extent that banks are agents seeking to maximize their profits by allocating their portfolio, the information they can get about the reliability is the of their customers main point of bank competition. Since one of the purposes of this analysis is to investigate the relevance of the financial structure and flow of profits for the firm's investment decision, and test it for a very large sample of firms, proper valuations of the market value of debt for firms not 153 operating on the stock and bond markets is not available, and, again, we argue that the book value should be an acceptable approximation. Even for what concerns physical capital it might be difficult to define a "market value" for capital coming from firm specific investments. In other words, the physical capital k might not have a market value at all if considered apart from its original production be unit, while it could extremely productive when employed within the firm's specific technology. The concept of market value of physical capital becomes even more ambiguous if we take into account firms. strategic interaction among Aoki and Leijohnufvud [1988] point out that the value of the endowment of physical capital of a firm is not independent from the endowment of physical capital of the competitors. Therefore, the market value of physical capital may become a very ambiguous concept in a context of imperfect competition, when the various competitors are able to adjust it according to strategic needs. These considerations could suggest that, to the extent that we are interested in explaining investments, the variation of the net book value of physical capital might be a reasonable approximation, since such variation reflects the cost of capital goods in the period under consideration. This if argument becomes stronger we think that for a very large sample like the of firms, one we are using here, there might not be a better approximation and measure for the investment expenditure. However, a price index of capital goods will be employed when the has to be capital stock expressed at different prices from those determining the book value. Moreover, distortions might be a problem definitions even for alternative of the empirical data. In 154 "market particular, empirical works where the so-called value" of defined (in physical capital is properly a non-securitized country), big firms, for data only refer to small samples of relatively which determined from the stock market, or detailed surveys (usually provided by the firms themselves, very rarely by independent institutions) are available. In the paper by Bernstein and Nadiri [1986], for example, the data set considered is a panel of forty- 1959-64. nine firms for the period In this regard it is very important to observe that the Bernstein and Nadiri model puts the financial firm emphasis on the market value of the assets of the and (as on the market value of R&D capital well as R&D expenditures), in based data by firms such as reported surveys on provided the themselves, and for what concerns the value of financial capital, by data. is stock market Moreover, there empirical evidence showing that the firms performing high levels of expenditure in R&D are financial funds raising a significant part of on the share market. As is well known, the share market is subject to several distortions bubbles, such as speculative and, according to the findings of the literature on share prices excess volatility (for example Shiller [1984], [1989]), they might not correctly reflect the net present value of dividends, and not, as a consequence, the proper value of firms. the assets of the Therefore, even the use of data obtained from share markets impose some form of prior restrictive assumption, by the is often violated empirical evidence, and for this reason it kind argued that the of approximation taken here might not contain distortion than the more elements of empirical works based on the Finally, it usual mainstream approach. might be worth mentioning in that firms whose data appears the Mediobanca survey, are usually 155 subject to some form of monitoring (such as official auditing), where criteria for data collecting might be no less rigorous and strict that the ones of the official statistical institutions. All the above considerations might lead us to conclude that, on one hand, the models emphasizing the "market values" of financial assets and physical capital might rely on suitable data only when they refer to very small and specific categories of firms, and that book on the other hand, the values of the variable considered in the be, in gearing ratio might our specific case, an acceptable approximation for the market values, given that any of these choices contains a degree of approximation. Since the data used here mainly concern firms not issuing securities, if we define "k" and "B" at their book values, then the leverage ratio, with regard to the balance sheet constraint 9, will be defined as B(t) Q(t) = (29 E(t) + R(t) Since the balance sheet constraint 9 holds at any moment in time, if k(t) and B(t) are defined at book values, the value of the be defined E(t)+R(t), own capital will as again at book value. The variable profits have been calculated, on the basis of Mediobanca data, as the value added minus the labour costs. Other details, such as the use of appropriate deflators for the prices of in capital goods are explained the appendix. The empirical results in the will be considered next section. 156 3. The empirical results The tables enclosed in the appendix contain some estimations made by following the technique of unbalanced panel data. The data have been obtained from the "Mediobanca" sample and refer to `three sectors: the chemical sector (tables 1-8), the electronic sector (tables 9-20) and the clothing sector (21-32). We start with equation 28 for the empirical specification of the investment function. Since in all of the three sectors the coefficient of the lagged dependent variable is close to one, the parameter restriction ß3=1 has been tested. Such restriction would yield equation 28' as an empirical specification : The usual F-test on parameter restrictions has not been employed in this case, for two main reasons. First of all, in the case of panel data, it is necessary to make use of some testing procedure robust to heteros Ce. dasticity, and the usual F-test based on the residual sum of squares is not robust to heterosCe: dasticity. Secondly, as it is well known, in the case of instrumental variables, the power of the test depends on the fact of having good instruments. In our case, looking at the unrestricted model (equation 28) the instruments for u/k and ln(ß) are their respective lagged variables, while for the lagged dependent value (i. e. lnIt-i) the instrument is a variable here defined as "LNKHLAG". This to lagged variable corresponds the value of the book value of the physical capital, gross of the accumulated depreciation (instrument for "LAGLOGIN", as defined below). -This value includes all of the installation costs of purchase and of all of the pieces of physical have become capital that not obsolete and are still in use in the 157 production process of the firm. Those which become obsolete earlier than expected are liquidated or eliminated from the balance sheet. The operation of liquidation originates an atypical or non-operative profit or loss. In this sense, the variable LNKHLAG can be thought of as the cumulated sum of the "purely technological" adjustment cost of investment implemented in the past and still in use in the firm. The kind of instrument employed for the lagged dependent variable is different from the one employed for the other regressors. Therefore, in contrasting the residual sum of squares of the restricted and unrestricted models, not only is there a problem of power of the test, but also a problem of homogeneity of the set of instruments employed. In fact, the "atypical" instrument LNKHLAG is only employed in the unrestricted model. Therefore, in order to test, for the parameter restriction 3=l, the following procedure has been implemented. First of all, the following regression has been implemented: U In It = PO + ßl - P2'ln(i2) + (1-ß3)"1n It-1 (28" , k Secondly, using the white (robust to heteroschedasticity) t- statistics the following null hypothesis has been tested: HO : (1-133) =0 In case the null hypothesis is rejected, equation 28 is still best regarded as the one. In case the null hypothesis is not 28' rejected, then equation is adopted as an empirical specification. However, the implications of the theoretical model here adopted require that if the parameter restriction ß3=1 holds, 28' then, in equation the regressors have to be non significant, as explained in the previous section. IL 158 It might be interesting to note that the empirical specification 28 would be likely to yield satisfactory results even if the "Kaleckian" interpretation of investment behaviour (such as formulated, for example, in Henley [1990]) is true. In fact, although the Kaleckian theory of investments is formulated in aggregated terms, it argues that the investments depend positively on the rate of profits and on the non-utilized production capacity. If one accepted that the leverage ratio could capture the effects of an unexpected negative shock in the profits, then it could be argued that the leverage ratio is correlated with the "unexpectedly non- utilized" production capacity. Another way of seeing a connection with the "Kaleckian" investment theory lies in the implications of the "deep pocket argument", or, in other words, the signalling use of the financial structure, mentioned in the introductory section. According to this approach, the signalling use of the financial structure leads the incumbent firms to choose a financial structure too expensive for the potential entry, in order to deter entry., Furthermore, by observing that the higher the level of investments higher at time t-1, the is likely to be the "non-utilized" (especially production capacity in the presence of an "entry- deterrence" use of investment), it could be observed that, loosely speaking, the explanatory power of the "non-utilized" production be jointly capacity could captured by the variables lnIt-1 and ln(n), while the rate of profits on physical capital is an investment explanatory cause of already present in the Kaleckian formulation. have been The estimations run using the package DPD (a routine by Arellano of Gauss developed and Bond [1988]). 159 Since the years employed for the estimations, for what concerns Italy, are in general regarded as years of uniform economic growth without any particular shock, the estimations for all of the equations have been implemented both with and without dummies. Appendix 1 contains the definition of the variables employed in the estimates and the tables with the estimates. Appendix 2 contains the list of all of the data for all of the firms and a detailed description of the way the data have been processed. The joint significance of the variables is assessed on the basis of the Wald test of joint significance, while the individual significance of the variables is assessed on the basis of the t- statistics. Table 1 shows the estimation of equation 28 for the chemical sector without annual dummies. The variables are jointly significant (at the level of confidence of 95%), and seem to be also individually significant, although lnmu is less significant than the others (the null hypothesis is only rejected with a level of 61 in the "one-step confidence of . estimates with robust test 2 statistics"). Table only shows some descriptive statistics and the asymptotic variance matrices. Table 3 shows the estimate for dummies, equation 28 with the time which are only significant with a level of confidence of 0.75. Apart from the dummies, both the value their of the coefficients and significance are analogous to the ones 1, shown in table representing equation 28 without dummies. Table 4 again shows some descriptive statistics and the asymptotic variance matrices of the equation estimated in table 3. 160 Having noticed that in equations 1 and 3 the coefficient for the lagged dependent variable (lnIt-1) is close to one, the parameter restriction ß3=1 for equation 28 has been tested. For this 28 purpose, equation has been reparametrized in the form of equation 28", where the null hypothesis H0: (1-ß3)=0 has been tested. Tables 5 28" and 7 show the estimates of equation without time dummies and with time dummies respectively. For the reasons explained at the beginning of this section, the test implemented for this purpose is the robust t-statistics on the significance of the coefficient of lnIt-i. Both in the case of table 5 and table 7, the null hypothesis H0: (1-ß3)=0 is rejected at the level of confidence of 0.99. This better means that the equations that describe the behaviour of investments for the firms of the chemical sector are the ones of table 1 and 3, reported as follows (the numbers in brackets refer to the robust t-statistics, and the definitions of the variables are 1): reported in appendix Chemical sector: estimate without time dummies. 777427 +. 922032 laglogin 001551 174732 lnmu login = . +. prorat -. (2.98282) (29.579329) (8.337532) (-. 871038) Wald (robust) test of joint significance = 1100.341505 739 Rz= . Chemical sector: estimate with time dummies. 850922 922766 laglogin 001655 login = . +. +. prorat -. 180182 lnmu + (3.238360) (29.868663) (8.978494) (-. 899436) 034625 D89 200129 D90 - . - . (-. 383689) (-1.796919) Wald (robust) test of joint significance - 1071.366471 Wald (robust) test of joint significance of time dummies = 3.419619 743 R2= . 161 Tables 9 and 11 show the estimate of equation 28 for the electronics sector without time dummies, and with time dummies respectively. Table 10 shows some descriptive statistics and the asymptotic variance matrices for the estimates of table 9, while table 12 provides the same information for the estimates of table 11. The time dummies are only significant at the level of confidence of 0.75. The variables are jointly significant, but in this case some variables are individually less significant than in the estimates for the chemical sector. Both in the estimate of table 9 and 11, the variable lnmu is not significant, while the variable "prorat" is only significant at the level of confidence of 0.55 in table 9 and 0.61 in table 11. In this case also the null hypothesis H0: (l-R3)=0 has been tested following the same procedure as in the chemical sector. Table 13 shows the estimate for equation 28" without time dummies (while table 14 shows again the same descriptive statistics and asymptotic 28" variances) and equation 15 shows the estimates for equation with time dummies (and again table 16 shows the respective descriptive statistics and variance matrices). Both in table 13 and 15 HO is not rejected, although only at the level of confidence of 0.90. This leads to a reformulation of the model according to equation 28' (table 17 for the specification without time dummies and table 19 for the specification with time dummies, while tables 18 and 20 show the respective descriptive statistics and variance matrices). As beginning explained at the of this section and at the end of section 2.1, this result could be interpreted as a situation where, for the firms, the fact of deviating from the previously signalled behaviour determines a very high level of "non-technological" or "strategical" 162 adjustment costs of investments. In other words, the cost of being out of the (theoretically) optimal level of investments is neglectable compared to the cost of deviating from the (possibly collusive) level of investments previously signalled. This implies (in order that our theoretical framework should be consistent with the theory) that the regressors of equation 28' must be non- significant. This seems to be the case, since the Wald test of joint significance for tables 17 and 19 yield a very low value (excepting for the test of joint significance of the time dummies) suggesting that all of the regressors of equation 28' are in this case non- significant. In the case of the electronic sector, although the results do not contrast with the theory, they yield a set of do information behaviour equations that not contain very much on the "strategic" of the firms' investments, apart from the fact that the cost of deviating from the previously signalled level of investments seems to be very costly. This yields a situation where only the past level of investments is strongly significant for the explanation of the present level of investments, while a certain degree of for ambiguity is still present what concerns the other variables. However, a situation where the firms' behaviour is strongly affected by the rivalry and/or collusion among the different firms and by a large use of signals, seems to fit with the characteristics of the electronics sector. Electronics sector: estimate in levels without time dummies, 123611 997099 laglogin 013059 012975 loginv = . +. +. prorat -. lnmu (0.32456) (23.729008) (0.756941) (-. 080122) Wald (robust) test of joint significance = 656.280974 815 R2- . 163 Electronics sector: estimate in levels with time dummies. 246333 997998 laglogin 014465 001043 lnmu loginv = . +. +. prorat -. + (0.669812) (24.053351) (0.872424) (-. 006415) 219409 -. 178303 D89 -. D90 (-1.356358) (-1.840832) Wald (robust) test of joint significance = 725.725731 Wald (robust) test of joint significance of time dummies = 3.711439 818 R2= . Electronics sector: estimate with the parameter restriction ß3=0 and without time dummies. 066896 013419 013873 lnmu dogin = . +. prorat -. (1.46291) (1.464071) (-. 125758) Wald (robust) test of joint significance = 0.757294 012 R2= . Electronics sector: estimate with the parameter restriction ß3=0 and with time dummies. 228658 014714 001651 lnmu 178467 D89 dogin = . +. prorat -. -. + (2.06872) (0.825006) (-. 012991) (-1.246084) -. 219552 D90 (-2.025493) Wald (robust) test of joint significance = 0.841852 Wald test of joint significance of time dummies = 5.042572 032 R2= . A result that seems to contrast with the theory has been the the obtained in case of clothing sector. Table 21 shows the estimate of equation 28 without- time dummies, table 22 shows its descriptive statistics and asymptotic Table 23 variance matrices. shows the estimates of equation 28 with time dummies, and table 24 shows its descriptive statistics and asymptotic variance matrices. The regressors are jointly 164 significant, and the time dummies are significant at the level of confidence of 0.95, therefore the equation of table 23 is the most reliable. In this equation, the coefficient for the variable lnmu has a wrong sign, but the variable itself is not significant. However, the results of the estimate of table 23 seem to contradict the theoretical framework of this paper, since the coefficient of the lagged dependent variable is greater than one. The null hypothesis H0: (l-ß3)=0 has been tested in tables 25 and 27 by running regression 28" and looking at the robust t-statistics for the lagged dependent variable. Table 25 shows the estimate for equation 28" without time dummies, table 26 shows its relative descriptive statistics and variance matrices, table 27 shows the 28" estimate for equation with time dummies and table 28 shows its descriptive statistics and variance matrices. In the estimate of table 25, HO is not rejected at the level of confidence of 0.90, 27 while in table the situation is more ambiguous, since HO is not level rejected only at the of confidence of 0.80. If the null hypothesis is not rejected, and the estimate of equation 28' is implemented (table 29 for equation 28' without time dummies and table 31 for equation 28' with time dummies), the results do not the seem to be consistent with theoretical framework followed here, in the since the regressors estimates of equation 28' are jointly the level significant at of confidence of 0.95. In fact, as it has been pointed out at the beginning of this section and at the end of theory followed in section 2.1, the this paper implies that when the is null hypothesis not rejected, then in equation 28' the regressor in must not be significant, as the case of the electronics sector. Therefore, the results of the estimates for the clothing sector do 165 not seem to fit with the theory described in this paper, although a certain degree of ambiguity remains due to the fact that the null hypothesis in table 27 has not been rejected only at the level of confidence of 0.80. However, it might be interesting to note that theoretical framework of this paper, designed to describe the investment behaviour with imperfect competition, yields the worst results in the clothing sector, which is, among the three considered here, the closest to the perfectly competitive configuration. Since the time dummies turned out to be significant, while the null hypothesis H0: (l-ß3)=0 has not been rejected, with a slightly higher degree of ambiguity than in the case of the electronics sector, only the estimates of table 23 and 31 (i. e. equation 28 and 28" both with time dummies) are reported in what follows: Clothing sector: estimate'in levels with time dummies. laglogin 056679 107080 lnmu loginv = . 107130 +1.032241 +. prorat +. + (0.078584) (5.489028) (1.552927) (. 444283) 275820 -. 633829 D89 -. D90 (-2.302265) (-. 978829) Wald (robust) test of joint significance = 37.181121 Wald (robust) test of joint significance of time dummies = 6.854276 200 Rz= . clothing sector: estimate with the parameter restriction ß3=0 and dummies. with time 340421 055356 095473 lnmu 613535 dogin = . +. prorat . -. D89 + (1.86353) (5.579749) (. 599836) (-2.586558) -. 254475 D90 (-1.000071) Wald (robust) test of joint significance = 32.403434 Wald (robust) test of joint significance of time dummies = 7.494676 129 R2= . 166 We can conclude, therefore, that the empirical results seem to be consistent with the theoretical framework of this paper for what concerns the chemical and electronics sectors, while they show some inconsistency with the theoretical framework of this paper (apart from showing low explanatory power and very scarce significance of the regressors) for what concerns the clothing sector. In this last 29 sector, the estimates of tables and 31 show a certain degree of between dlogin (difference statistical correlation the term of the logs of the investments) and the term prorat (rate of profits on be interesting physical capital). It might to note that the worst in (clothing) results have been obtained a sector whose market (at least structure is much closer to the perfectly competitive one for what concerns the presence of technological know-how, barriers to entry and average size of the firms) than the other two sectors distant analyzed here. In this sense, the clothing sector is more from the theoretical framework employed, which justifies the lagged dependent presence of a variable on the basis of the effect in different of stability vs. modifications the conjectures of the firms operating in an imperfectly competitive context. 4. Conclusions. The purpose of this paper is to provide and test empirically a dynamic framework simplified where the decision concerning investments and the financial structure of the firm take place in the bankruptcy simultaneously, presence of and adjustment costs imperfect and with competition on the goods market. An important feature of this model is the presence of non-technological 167 adjustment costs of investments, determined by a deviation from the level of investments such as signalled at discrete intervals. In section 3 an empirical analysis has been implemented on the basis of the theoretical model presented in section 2. The empirical analysis is based on three samples of firms' data from three different industries: the chemical, the electronics and the clothing industries. The results, obtained through panel data techniques, seem to be consistent with the predictions of our theoretical framework for what concerns the chemical and the electronics sectors, while they show a few inconsistencies for what concerns the clothing sector, i. e. the sector whose market configuration seems to be closest to the perfectly competitive one. A possible explanation for such inconsistencies might be found in the simplifying assumptions made on the very complex relations between the cost of capital, the profitability of the firm, and the different strategic interactions among the competitors. A big number in industrial of theoretical contributions economics and in finance different point out many relations of causality and possible functional links among these variables. These issues, as well as an interpretation alternative of the investment behaviour, constitute the purpose of the analysis of the next chapter. 168 Bibliography of Chapter 5 Abel, A. B., Eberly, J. C., [1993] "A Unified Model of Investment Under Uncertainty", NBER Working Paper Series no. 4296 Aoki, M., Leijohnufvud, A. [1988] " The Stock-Flow Analysis of Investments" in Kohn and Tsiang: "Financial Constraints, Expectations and Macroeconomics", Oxford, Oxford University Press. Arellano, M., Bond, S. R., [1988] "Dynamic Panel Data Estimation Using DPD -a Guide for Users" Institute for Fiscal Studies Working Paper, no. 88/15 Benoit, J. -P., [1984], "Financially Constrained Entry in a Game with Incomplete Information" RAND Journal of Economics, 15, pp. 490- 499. Benston, G. J. [1982], "Accounting Numbers and Economic Values", Antitrust Bulletin, 21, pp. 161-215. Bernanke, B., [1983] "Non-Monetary effects of the Financial Crisis in the Propagation of the Great Depression", American Economic 73 257-276 Review, vol. pp. . Bernanke, B. e Blinder, A. (1988) "Is it Money or Credit, or Both, or Neither? Credit, Money, and aggregate Demand" American Economic Review: Papers and Proceedings, may; Princeton University, New Jersey Bernanke, B., Gertler, M. [1989] "Agency Costs, Net Worth, and Business Fluctuations", American Economic Review, vol. 79, pp. 14-31. Bernstein, J., I., Nadiri, M. I., [1986] "Financing and Investment in Plant and Equipment and Research and Development", in Peston, M. H., Quandt, R. E., "Prices, Competition and Equilibrium", Oxford, Phillips Allan Publishers. Bernstein, J. I., Nadiri, M. I. [1993], "Production, Financial Structure and Productivity Growth in U. S. Manufacturing", NBER Working Paper Series No. 4309 Blanchard, O. J., Lopez-de-Silanez, F., Shleifer, A., [1993], "What Do Firms Do With Cash Windfalls? ", NBER Working Paper series no. 4258 Blundell, R., Bond, S., Meghir, C. [1992], "Econometric Models of Company Investments", in Matyas, L. and Sevestre, P. (eds. ) "The Econometrics of Panel Data - Handbook of Theory and Applications", Kluwer Academic Publishers, Dordrecht, Holland. Bond, S., Meghir, C. [1992] "Dynamic Investment Models and the Firm's Financial Policy" no. 92-14, UCL Discussion Papers In Economics, Department of Economics, University College London. Brander, J. A., Lewis, T. R., [1986] "Oligopoly and the Financial 169 Structure - The Limited Liability Effect", American Economic Review, vol. 75, pp. 956-970. Brioschi, F., Buzzacchi, L., Colombo M. G., [1990], "Gruppi di imprese e Mercato Finanziario", Roma, La Nuova Italia Scientifica. Diamond, D., [1984] "Financial Intermediation and Delegated Monitoring", Review of Economic Studies, vol. 51, pp. 393-414. Diamond, D., [1989] "Reputation Acquisition in Debt Markets", Journal of Political Economy, vol. 97 pp. 828-862. Fazzari, S. M., Hubbard, G. R., Petersen, B. C. [1988], "Financing Constraints and Corporate Investment", Brooking Papers on Economic Activity, vol. 1, pp. 141-206. Frasca, F., Marotta, G., [1988], "La Ristrutturazione Finanziaria delle Grandi Imprese", in Atti del Seminario "Ristrutturazione Economica e Finanziaria delle Imprese", Roma, Banca d'Italia. Ganoulis, I., [1991], "Stock Market Prices and Investment when Financial Markets are not Strongly Efficient", Applied Economics Discussion Paper Series no. 128, Institute of Economics and Statistics, Oxford University. Gardener, E. P. M., (1991) "Nuove strategie per le banche europee", Banca Impresa, Societä Anno X, 1, aprile; 19-39; Bologna, I1 Mulino, Gardener, E. P. M. e Molyneux, P., (1990) Changes in Western European Banking; Unwin Hyman, London Gertler, M. [1988] "Financial Structure and Aggregate Economic Activity: An Overview", Journal of Money, Credit and Banking, vol. 20, pp. 559-588. Greenwald, B. C., Stiglitz, J. E., [1988] "Imperfect Information, Finance Constraints, and Business Fluctuations", in Kohn and Tsiang: "Financial Constraints, Expectations and Macroeconomics", Oxford, Oxford University Press. Greenwald, B. C., Stiglitz, J. E., [1989] "Financial Market Imperfections and Productivity Growth", NBER Working Paper n. 2945. Greenwald, B. C., Stiglitz, J. E., [1990a] "Asymmetric Information and the New Theory of the Firm: Financial Constraints and Risk Behaviour", NBER Working paper n. 3359. Greenwald, B. C., Stiglitz, J. E., [1990b] "Macroeconomic Models with Equity and Credit Rationing", NBER Working Paper, n. 3533. Greenwald, B. C., Stiglitz, J. E., Weiss, A., [1984] "Informational Imperfections in the Capital Market and Macroeconomic Fluctuations", American Economic Review, vol. 74, pp. 194-199. 170 Hayashi, F. [1982], "Tobin's Marginal q and Average q: A Neoclassical Interpretation", Econometrica, vol. 50, pp. 213-224 Henley, A., [1990], "Wages and Profits in the Capitalist Economy - The impact of monopolistic Power on macroeconomic Performance" Aldershot, U. K., Edward Elgar. Hörngren, L. [1985] "Regulatory Monetary Policy and Uncontrolled Financial Intermediaries". Journal of Money, Credit and Banking, vol 17, pp. 203-219. Hsiao, C. [1986], "Analysis of Panel Data", Cambridge, U. K., Cambridge University Press. Leland, H., Pyle, D., [1977) "Information Asymmetries, Financial Structure and Financial Intermediation", The Journal of Finance, vol. 32, pp. 371-398. Martin, S. [1993], "Advanced Industrial Economics", Oxford, Blackwell. Mayer, C. (1989) "The Influence of the Financial System on the British Corporate Sector", Mimeo prepared for the conference "The Separation of Industry and Finance and the Specialization of Financial Intermediaries" at Universitä Bocconi, Milano Mayer, C., [19921 "Corporate Finance", mimeo, forthcoming in The New Pal grave Dictionary of Economics and Finance Mayer, C. [1993] "Ownership", mimeo, University of Warwick Myers, S. C. [1984] "The Capital Structure Puzzle", The Journal of Finance, vol. 39, pp. 575-592. [1984], "Corporate Myers, S. C., Majluf, N. S. Financing and Investment Decisions when Firms Have Information that Investors Do Not Have", Journal of Financial Economics, vol. 13, pp. 187- 221. Okun, A. M., [1981] "Prices and Quantities: A Macroeconomic Analysis", 1981 Washington, D. C., Brookings Institution. [1989a], "Financial 'Deep Poitervin, M. Signalling and the Pocket' Argument", Rand Journal of Economics, 20, pp. 26-40. [1989b] "Collusion Poitervin, M., and Banking Structure of a Duopoly", Canadian Journal of Economics, XXII, no. 2, pp. 263- 277. "Strategic poitervin, M. [1990], Financial Argument", international journal of Industrial Organization", 8, pp. 499-518. [1987], "Rational Expectations, Precious, M. Non-Market Clearing and Investment Theory", Oxford, Clarendon Press. 171 Ross, S. [1977] "The Determination of Financial Structure: The Incentive Signalling Approach", Bell Journal of Economics, vol. 8, pp. 23-40. Shiller, R. [1981], "Do Stock Prices Move Too much to Be Justified by Subsequent Changes in Dividends? " American Economic Review, vol. 71, pp. 421-436 Shiller, R. [1984], "Stock Prices and Social Dynamics", Brooking Papers on Economic Activity vol. 2, pp. 457-510. Shiller, R. [1989] "Market Volatility", Cambridge, Massachussets, M. I. T. Press. Summers, L., H. [1986], "Do Market Prices Accurately Reflect Fundamental Values? ", The Journal of Finance, vol. 41, pp. 591- 601. Stiglitz, J. E., [1974] "On The Irrelevance of Corporate Financial policy", American Economic Review, vol. 64, pp. 851-866. Telser, L. G., "Cutthroat Competition and the Long Purse" Journal of Law and Economics, 20,1989, pp. 26-40. 172 APPENDIX 1 In what follows the symbols of the Variables Included in the Estimations and in the equations are reported. The first section refers to the instruments, the second 1. INSTRUMENTS: CONST = (constant) intercept term; LNKHLAG = lagged value of the book value of the physical capital, depreciation (instrument gross of the accumulated for "laglogin", as defined below). This value includes all of the costs of purchase and installation of all of the pieces of physical capital, that have not become obsolete and are still in use in the production process of the firm. Those which become obsolete earlier than expected are liquidated or eliminated from the balance sheet. The operation of liquidation originates an extraordinary profit or loss. In this lnkhlag be sense, the variable can thought of as the cumulated sum investment of the "purely technological" adjustment cost of implemented in the past and still in use in the firm. PRORATLA = lagged value of "prorat", as defined below; LNMULA=log of the lagged value of (l+n), (where the debt only include log term financial debt). (l+A) has been employed instead of p, since 0 is often null, and could not have been calculated in logs. 2. DEPENDENT VARIABLES LOGINN = lnI(t) as defined in equation 37. DLOGINV = lnI(t) - lnI(t-1) 173 D89 = dummy variable for year 1989. D90 = dummy variable for year 1990. 3. REGRESSORS: LAGLOGIN = lnI(t-1) as defined in equation 37; lnmu = ln(l+Q) u(t) PRORAT = k however, u(t)/k, determines some problems of approximation. u(t) is a flow variable determined between t-1 and t, while k has to be defined either at time t or at time t-l. Its price, again, has to be defined either at time t or t-1. Any choice would contain some degree of approximation. The choice made here, analogous to the one made by Bernstein and Nadiri [1986], is the following: var. prof. (t) PRORAT = pk(t)k(t-1) where VAR. PROF. - variable profits, i. e., from Mediobanca data, the difference between the value added and the labour cost. The implicit here simplifying assumption is that it is the capital stock at time t-l that contributes to determine the variable profits at time t. However, the capital stock, although considered at time t-l for the has sake of simplification, to be valuated at a price level the time calculated at same when the variable profits are calculated, i. e. time t. 174 deflator Pk = implicit price of capital goods (source DATASTREAM services at the University of Warwick, on the basis of OECD data); in particular the data are the following: year price at time t/ price at time t-1 1986 1.05688 1987 1.06135 1988 1.05066 1989 1.06159 1990 1.0549 1991 1.041 4. "RAW" DATA AND VARIABLES: ACC. D. (t) = Accumulated depreciations at time "t"; DEPR(t) = Depreciations at time "t"; I(t) = gross investment (INV. in the tables of the data), defined as follows: (t) I(t) = K(t) - ACC. D. - K(t-1) + ACC. D. (t-1) + DEPR(t) (VAR. u(t) = variable profits PROF. in the tables of the data) defined as the difference between the value added (V. ADD. ) and the labor cost (LAB. C. ); (Book EQ. = Equities Value); RES. = balannce sheet reserves and accumulated profits; L. T. F. D. = long term financial debt; )/L. T. F. D. this has mu = (EQ. +RES. ; ratio been earlier defined as 0; the variable lnmu is not actually the log of mu, but its proxy. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 1 CHEMICAL SECTOR D. P. D. RESULTS LEVELS IV Number of firms: 124 Sample period is 1988 to, 1990 Observations: 274 Degrees of freedom: 270 Dependent variable is: loginv Instruments used are: CONST lnkhlag proratla lnmula ONE-STEP ESTIMATES RSS = 136.145185 TSS = 521.925911 Estimated sigma-squared (levels) = 0.504241 Wald t est of joint significance: 713.365013 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.777427 0.287009 2.708721 0.006754 laglogin 0.922032 0.034794 26.499427 0.000000 0.001551 0.000695 prorat 2.230373 0. "025723 lnmu -0.174732 0.209604 -0.833627 0.404491 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -1.808 [ 89 ] Test for second-order serial correlation: -2.434 [ 61 3 ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 1100.341545 df =3 Var Coef Std. Error T-Stat P-Value CONST 0.777427 0.260635 2.982820 0.042856 laglogin 0.922032 0.031171 29.579329 0.000000 prorat 0.001551 0.000186 8.337532 0.000000 lnmu -0.174732 0.200602 -0.871038 0.383734 for first-order Robust test -serial correlation: -1.084 C 89 ] for Robust test second-order serial correlation: -1.621 C 61 ] Estimated serial correlation matrix 1.000 0.013 1. CG00 1.000 -0.247 -0.266 Number of observations available to sample covariances 86 76 97 61 74 91 identified two-step Model just - estimates and one-step estimates coincide ------TABLE 2 CHEMICAL SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max loginv 8.29173 1.38268 3.80666 11.43889 laglogin 8.18781 1.39863 2.94444 11.23452 prorat 5.60625 64.72126 -0.54212 1070.34302 lnmu 0.25076 0.25699 0.00000 1.52245 Correlation Matrix loginv laglogin prorat lnmu 1.00 0.86 1.00 -0.07 -0.15 1.00 0.11 0.12 1.00 , -0.06 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x1OOO00) CONST 1aglogin prorat lnmu 8237.406 -970.297 121.065 -3.612 0.380 0.048 -352.984 -92.076 0.927 4393.389 Robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu 6793.052 -794.295 97.166 -3.445 0.374 0.003 -391.501 -42.124 0.625 4024.107 ...... TABLE 3 CHEMICAL SECTOR D. F. D. RESULTS LEVELS IV Number of firms: 124 Sample period is 1988 to 1990 Observations: 274 Degrees of freedom: 268 Dependent variable is: loginv Instruments used are: CONST lnkhlag proratla lnmula TIM DUMS ------ONE-STEP ESTIMATES RSS = 134.188713 TSS =5 21.925911 Estimated sigma-squared (levels) = 0.500704 Wald test of joint significance: 718.765907 df = 3 Wald test - it sig of time dums: 4.080675 df = 2 Var Coef Std. Error T-Stat P-Value CONST 0.850922 0.291546 2.918658 0.003515 laglogin 0.922766 0.034681 26.607521 0.000000 prorat 0.001655 0.000696 2.378346 0.017390 lnmu -0.180182 0.208825 -0.862841 0.388225 D69 -0.034625 0.104818 -0.330339 0.741144 D90 -0.200129 0.106813 -1.873648 0.060979 NOTE: Standard errors and test statistic s not robust to heteroskedasticity Test for first-order serial correlation : -1.813 C 89 7 Test for second-order serial correlation: -2.266 C 61 3 ------ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 1071.366471 df = 3 Wald test - it sig of time dums: 3.419619 df = 2 Var Coef Std. Error T-Stat P-Value CONST 0.850922 0.262763 3.238360 0.001202 laglogin 0.922766 0.03.0894. 29.868663 0.000000 prorat 0.001655 0.000184 8.978494 0.000000 inmu -0.180182 0.200328 -0.899436 0.368421 D89 -0.034625 0.090243 -0.383689 0.701209 D90 -0.200129 0.111374 -1.796919 0.072348 Robust t est for first-order serial correlation: -1.112 C 89 3 Robust t est for second-order serial correlation: -1.629 C 61 3 Estimated serial correlation matrix 1.000 0.002 1.000 1.000 -0.233 - 0.260 Number of observations available to sample covariances 86 76 97 6i 74 91 Model just identified - two-step estimates and one-step estimates coincide ------TABLE 4 CHEMICAL SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max loginv 8.29173 1.38268 3.80666 11.43889 laglogin 8.18781 1.39863 2.94444 11.23452 prorat 5.60625 64.72126 -0.54212 1070.34302 lnmu 0.25076 0.25699 0.00000 1.52245 Correlation Matrix loginv laglogin prorat lnmu 1.00 0.86 1.00 -0.07 -0.15 1.00 0.11 0.12 -0.06 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100040) CONST laglogin prorat lnmu D89 D90 8499.897 -959.771 120.275 -3.446 0.3+84 0.048 -322.892 -91.582 0.931 4360.777 -574.327 0.062 -0.013 -34.140 1098.675 -467.075 -12.728 -0.607 -41.846 582.659 1140.894 Robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu D89 D90 6904.459 -773.526 95.445 -3.138 0.359 0.003 -502.517 -48.320 0.404 4013.145 -258.264 -24.767 -0.030 93.552 814.384 -604.220 9.038 -0.592 389.434 434.822 1240.407 ttt+t+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 5 CHEMICAL SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 124 Sample period is 1988 to 1990 Observations: 274 Degrees of freedom: 270 Dependent variable is: dloginv Instruments used are: CONST lnkhlag proratla lnmula ------ ONE-STEP ESTIMATES RSS = 136.145185 TSS = 147.403211 Estimated sigma-squared (levels) =-0.252121 Wald test of joint significance: 20.239797 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.777427 0.204536 3.800901 0.000144 laglogin -0.077968 0.024736 -3.152046 0.001621 prorat 0.001551 0.000683 2.269625 0.023230 lnmu -0.174732 0.159402 -1.096172 0.273004 NOTE: Standard errors and te st statistics not robust to heteroskedasticity Test for first-order serial correlation: -1.882 C 89 ] Test for second-order serial correlation: -2.448 C 61 ] ------ ONE-STEP ESTIMATES WITH ROBU ST TEST STATISTICS Wald test of joint signif icance: 191.495887 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.777427 0.286724 2.711415 0.006700 laglogin -0.077968 0.034015 -2.292135 0.021898 prorat 0.001551 0.000195 7.937991 0.000000 lnmu -0.174732 0.172805 -1.011148 0.311946 Robust test for first-order serial correlation: -1.343 C 89 ] for Robust test second-order serial correlation: -1.624 C 61 3 Estimated serial correlation matrix 1.000 0.013 1.000) 1.000 -0.247 -0.266 Number of observations available to sample covariances 86 76 97 61 74 91 Model just identified - two-step estimates and one-step estimates coincide ------TABLE 6 CHEMICAL SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv 0.10392 0.73481 -3.58269 2.76505 laglogin 8.18781 1.39863 2.94444 11.23452 prorat 5.60625 64.72126 -0.54212 1070.34342 lnmu 0.25076 0.25699 0.00000 1.522245 Correlation Matrix dloginv laglogin prorat lnmu i. Ut) -0.28 1.00 0.16 -0.15 1.00 -0.02 O. 12 -0.06 1.00 ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu 4183.560 -491.340 61.186 -2.338 0.248 0.047 -287.520 -41.739 0.582 2540.893 Robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu 8221.048 -960.425 115.705 -3.987 0.442 0.004 -356.808 -30.050 0.620 2986.166 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 7 CHEMICAL SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 124 Sample period is 1988 to 1990 Observations: 274 Degrees of freedom: 268" Dependent variable is: dloginv Instrumen ts used are: CONST lnkhlag proratla lnmula TIM DUMS ------ONE-STEP ESTIMATES RSS = 134.188713 TSS =1 47.403211 Estimated sigma-squared (levels) _, 0.250352 Wald test of joint si gnificance: 21.120679 df = 3 Wald test - it sig of time dums: 18.490814 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.850922 0.217426 3.913627 0.000091 laglogin -0.077234 0.024657 -3.132315 0.001734 0.015574 prorat 0.001655 0.000684 2.418750 lnmu -0.180182 0.158791 -1.134716 0.256494 D89 -0.034625 0.124695 -0.277681 0.781257 D90 -0.200129 0.106666 -1.876220 0.060625 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -1.888 C 89 ] Test for second-order serial correlation: -2.279 C 61 ] ------ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 212.232895 df = 3 Wald test - it sig of time dums: 9.381470 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.850922 0.291524 2.918873 0.003513 laglogin -0.077234 0.033557 -2.301539 0.021361 prorat 0.001655 0.000186 8.880043 0.000000 lnmu -0.180182 0.173496 -1.038541 0.299018 D89 -0.034625 0.090164 -0.384027 0.700958 D90 -0.200129 0.119150 -1.679641 0.093027 Robust test for first-order serial correlation: -1.367 [ 89 ] Robust test for second-order serial correlation: -1.646 [ 61 ] Estimated serial correlation matrix 1.000 0.002 1.000 1.000 -0.233 -0.260 Number of observations available to sample covariances 86 76 97 61 74 91 Model just identified - two-step estimates and one-step estimates coincide ------TADLE 8 CHEMICAL SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv 0.10392 0.73481 -3.58269 2.76505 laglogin 8.18781 1.39863 2.94444 11.23452 prorat 5.60625 64.72126 -0.54212 1070.34302 lnmu 0.25076 0.25699 0.00000 1.52245 Correlation Matrix dloginv laglogin prorat lnmu 1.00 -0.28 1.00 0.16 -0.15 1.00 -0.42 C). 12 -4.06 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) " CONST laglogin prorat lnmu D89 D90 4727.385 -493.377 60.797 -2.265 0.249 0.047 -259.151 -41.429 0.578 2521.450 -882.083 9.873 0.213 -37.851 1554.877 -598.587 3.018 -0.544 -27.834 597.461 1137.768 Robust AVM of one-step estimates (x100000) CONST laglogin prorat Inmu D89 D90 8498.642 -942.051 112.610 -3.579 0.405 0.003 -407.754 -41.518 0.455 3010.079 -328.169 -15.461 0.037 75.519 812.951 -1007.960 48.375 -0.384 353.130 414.646 1419.674 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 9 ELECTRONICS SECTOR D. P. D. RESULTS LEVELS IV Number of firms: 73 Sample period is 1988 to 1990 Observations: 163 Degrees of freedom: 159 Dependent variable is: loginv Instruments used are: CONST lnkhlag proratla lnmula ------ ONE-STEP ESTIMATES 72.782900 TSS ,392 RSS . 534262 (levels) 457754 Estimated sigma-squared =0 . Wald t est of joint significance: 6 99.198213 df =3 Var Coef Std. Error T-Stat P-Value CONST 0.123611 0.378417 0.326654 0.743930 laglogin 0.997099 0.042020 23.729008 0.000000 prorat 0.013059 0.013967 0.935046 0.349765 lnmu -0.012975 0.154067 -0.084216 0.932885 NOTE: Sta ndard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -3.564 C 52 ] Test for second-order serial correlation: 1.789 E 38 ] ------ ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 656.280794 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.123611 0.380848 0.324569 0.745508 laglogin 0.997099 0.043560 22.889978 0.000000 prorat 0.013059 0.017253 0.756941 0.449085 lnmu -0.012975 0.161940 -0.080122 0.936140 Robust test for first-order serial correlation: -2.076 C 52 ] Robust test for second-order serial correlation: 1.699 C 38 ] E stimated serial correlation matrix 1.000 -0.230 1.000 0.259 -0.485 1.000 Number of observations available to sample covariances 50 42 52 '8 48 61 identified two-step Model just - estimates and one-step estimates coincide ------TABLE 10 ELECTRONICS SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max loginv 8.81888 1.55662 5.84644 13.92139 laglogin 8.69317 1.54594 5.48894 13.92139 prorat 2.50357 4.82378 -1.41262 46.03984 lnmu 0.41392 0.42540 0.00000 1.91044 Correlation Matrix loginv laglogin prorat lnmu 1.00 0.91 1.00 -0.33 -0.38 1.00 0.22 0.22 -0.19 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu 14319.978 -1541.109 176.570 -264.279 23.414 19.507 47.383 -126.749 28.744 2373.667 Robust AV M of one-s tep estimates (x300000) CONST laglogin prorat lnmu 14504.557 -1613.254 189.752 -323.165 27.247 29.766 311.210 -154.356 61.602 2622.449 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 11 ELECTRONICS SECTOR D. P. D. RESULTS LEVELS IV Number of firms: 73 Sample period is 1988 to 1990 Observations: 163 Degrees of freedom: 157 Dependent variable is: loginv Instruments used are: CONST lnkhlag proratla lnmula TIM DUMS ------ONE-STEP ESTIMATES RSS = 71.356743 TSS = 392.534262 Estimated sigma-squared (levels) = 0.454502 Wald test joint of significance: 704.168013 df =3 test it Wald - sig of time dams: 3.163735 df =2 Var Coef Std. Error T-Stat P-Value CONST 0.246333 0.383129 0.642951 0.520256 laglogin 0.997998 0.041849 23.847676 0.000000 prorat 0.014465 0.013781 1.049584 0.293909 lnmu -0.001043 0.153039 -0.006814 0.994564 D89 -0.176303 0.133653 -1.334077 0.182179 D90 -0.219409 0.128708 -1.704703 0.088250 NOTE: Standard errors and test statistics not robust to heteroskedasticity for first-order Test serial correlation: -3.540 C 52 7 Test for second-order serial correlation: 1.703 38 ] ------ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS test joint Wald of significance: 725.725731 df =3 it Wald test - sig of time dums: 3.711439 df =2 Var Coef Std. Error T-Stat P-Value CONST 0.246333 0.367765 0.669812 0.502978 laglogin 0.997998 0.041491 24.053351 0.000000 prorat 0.014465 0.016580 0.872424 0.362977 lnmu -0.001043 0.162555 -0.006415 0.994882 D69 -0.178303 0.131457 -1.356358 0.174985 D90 -0.219409 0.119190 -1.840832 0.065646 test for first-order Robust serial correlation: -2.024 C 52 ] Robust test for second-order serial correlation: 1.688 38 ] Estimated serial correlation matrix 1.000 -0.222 1.000 0.257 -0.492 1.000 Number of observations available to sample covariances 50 42 52 38 48 61 just identified two-step Model - estimates and one-step estimates coincide ------TABLE 12 ELECTRONICS SECTOR DESCRIPTIVE STATISTICS Variable Mean ' Std Dev Min Max loginv 8.81888 1.55662 5.84644 13.92139 laglogin 8.69317 1.54594 5.48894 13.92139 prorat 2.50357 4.82378 -1.41262 46.03984 lnmu 0.41392 0.42540 0.00000 1.91044 Correlation Matrix loginv laglogin prorat Inmu 1.00 0.91 1.00 -0.33 -0.38 1.00 0.22 0.22 -0.19 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100400) CONST laglogin prorat lnmu D89 D90 14678.793 -1518.178 175.133 -260.795 23.003 18.992 43.610 -127.023 26.264 2342.093 -811.273 -14.631 3.526 44.791 1786.304 -824.376 -12.352 3.431 29.789 911.825 1656.582 Robust AV M of one-step estimates (x100000) CONST laglogin prorat lnmu D89 D90 13525.079 -1451.431 172.150 -279.743 24.206 27.489 287.264 -165.761 51.855 2642.417 -945.312 7.512 2.134 346.529 1728.099 -593.291 -12.755 -30.465 132.884 725.937 1420.635 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 13 ELECTRONICS SECTOR D. R. D. RESULTS FIRST DIFFERENCES IV Number of firms: 73 Sample period is 1988 to 1990 Observations: 163 Degrees of freedom: 159 Dependent variable is: dloginv Instruments used are: CONST lnkhlag proratla Inmula ONE-STEP ESTIMATES 72.782900 TSS 73 RSS = = . 795977 (levels) Estimated sigma-squared ="0 . 228877 Wald t est of joint significance: 2.579555 df =3 Var Coef Std. Error T-Stat P-Value CONST 0.123611 0.269935 0.457930 0.647003 laglogin -0.002901 0.029827 -0.097272 0.922510 prorat 0.013059 0.010154 1.286158 0.198388 inmu -0.012975 0.112252 -0.115587 0.907980 NOTE: Sta ndard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -3.464 [ 52 ] Test for second-order serial correlation: 1.789 [ 38 3 ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 0.829524 df =3 Var Coef Std. Error T-Stat P-Value CONST 0.123611 0.272793 0.453133 0.650453 laglogin -0.00901 0.029697 -0.097698 0.922172 prorat 0.013059 0.019252 0.678333 0.497560 lnmu -0.012975 0.125432 -0.103442 0.9176 1 Robust test for first-order serial correlation: -1.940 C 52 ] Robust test for second-order serial correlation: 1.710 [ 38 3 Estimated serial correlation matrix 1.000 1.000 -0.230 0.259 -0.485 1.000 Number of observations available to sample covariances 50 42 52 38 48 61 Model just identified - two-step estimates and one-step estimates coincide ------TABLE 14 ELECTRONICS SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv 0.12571 0.67493 -2.26800 2.44715 laglogin 8.69317 1.54594 5.48894 13.92139- prorat 2.50357 4.82378 -1.41262 46.03984 lnmu 0.41392 0.42540 0.00000 1.91044 Correlation Matrix dloginv la glogin prorat lnmu 1.00 -0.20 1.00 0.11 -0.38 1.00 -0.01 0.22 -0.19 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu 7286.503 -779.971 88.964 10.310 -148.707 13.079 86.732 -75.324 20.274 1260.060 Robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu 7441.586 -769.965 88.190 -257.883 16.173 37.065 -180.341 -68.071 93.711 1573.325 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 15 ELECTRONICS SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 73 Sample period is 1988 to 1990 Observations: 163 Degrees of freedom: 157 Dependent variable is: dloginv Instruments used are: CONST 1nkhlag proratla lnmula TIM DUMS ------ONE-STEP ESTIMATES - RSS = 71.356743 TSS = 73.795977 Estimated sigma-squared (leve ls) = 0.227251 Wald test of joint significance: 3.009473 df = 3 Wald t est - it sig of time dums: 3.191340 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.246333 0.281894 0.873849 0.382200 laglogin -0.002002 0.029695 -0.067434 0.946236 0.144700 prorat 0.014465 0.009917 1.458512 lnmu -0.001043 0.111320 -0.009367 0.992526 D89 -0.178303 0.158610 -1.124160 0.260945 D90 -0.219409 0.128458 -1.708031 0.087631 NOTE: Sta ndard errors and tes t statistics not robust to heteroskedasticity Test for first-order serial correlation: -3.448 C 52 ] Test for second-order serial correlation: 1.703 C 38 ] ------ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint signifi cance: 0.900979 df = "Ir Wald test - it sig of time dums: 4.974831 df = 3 Var ' Coef Std. Error T-Stat P-Value CONST 0.246333 0.266984 0.922650 0.356190 laglogin -0.002002 0.028929 -0.069219 0.944815 prorat 0.014465 0.018432 0.784743 0.432604 lnmu -0.001043 0.128440 -0.008119 0.993522 089 -0.178303 0.143126 -1.245774 0.212848 D90 -0.219409 0.108533 -2.021591 0.043219 Robust test for first-order serial cor relation: -1.938 C 52 ] Robust test for second-order serial cor relation: 1.715 38 ] Estimated serial correlation matrix 1.000 1.000 -0.222 0.257 -0.492 1.000 Number of observations available to sample covariances 50 42 52 38 48 61 Model just identified - two-step estimates and one-step estimates coincide ------TABLE 16 ELECTRONICS SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv 0.12571 0.67493 -2.26800 2.44715 laglogin 8.69317 1.54594 5.48894 13.92139 prorat 2.50357 4.82378 -1.41262 46.03984 1nmu 0.41392 0.42540 0.00000 1.91044 Correlation Matrix' dloginv la glogin prorat Inmu 1.00 -0.20 1.00 0.11 -0.38 1.00 -0.01 0.22 -0.19 1.00 ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (xl00000) CONST laglogin prorat lnmu D89 D90 7946.438 -772.217 88.180 -140.582 12.647 9.835 147.064 -77.374 17.745 1239.207 -1213.429 -6.157 - -1.988 -4.168 2515.715 1.185 -1015.105 14.452 -46.751 929.083 1650.134 Robust AV M of one-step esti mates (x1 00000) CONST laglogin prorat lnmu D89 D90 7128.069 -704.159 83.691 -234.033 14.787 33.974 -26.904 -83.186 83.935 1649.682 -859.015 -9.891 18.656 18.120 2048.515 -276.275 -30.589 -16.186 -48.255 415.799 1177.942 t+ t+ TrTrrttttttttttttttt++++++++t+t+++++++++++t++++++ttt+tt++++++t+tt++tt++. TABLE 17 ELECTRONICS SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 73 Sample period is 1988 to 1990 Observations: 163 Degrees of freedom: 160 Dependent variable is: dloginv Instruments used are: CONST proratla Inmula ONE-STEP ESTIMATES RSS = 72.937946 TSS = 73.795977 Estimated sigma-squared (levels) _-0.227931 Wald test of joint significance: 2.489874 df =2 Var Coef Std. Error T-Stat P-Value CONST 0.097862 0.066896 1.462909 0.143492 prorat 0.013419 0.009165 1.464071 0.143175 lnmu -0.013873 0.110314 -0.125758 0.899923 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -3.470 C 52 ] Test for second-order serial correlation: 1.795 C 38 ] ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significances 0.757294 df -2 Var Coef Std. Error T-Stat P-Va 1 ue CONST 0.097862 0.085144 1.149370 0.250404 prorat 0.013419 0.018641 0.719857 4.471613 lnmu -0.013873 0.124524 -0.111407 0.911294 Robust test for first-order serial correlation: -1.915 C 52 ] Robust test for second-order serial correlations 1.711 1 38 3 Estimated serial correlation matrix 1.000 -0.231 1.000 0.260 -0.485 1.000 Number of observations available to sample covariances 50 42 52 Z.8 48 61 two-step Model just identified - estimates and one-step estimates coincide ------TABLE 18 ELECTRONICS SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv 0.12571 0.67493 -2.26800 2.44715 prorat 2.50357 4.82378 -1.41262 46.03984 lnmu 0.41392 0.42540 8.00008 1.91044 Correlation Matrix dloginv prorat Inmu 1.00 C). 11 1.00 1.00 -0.01 -0.19 ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST prorat lnmu 447.501 -33.678 8.400 -576.415 30.114 1216.908 Robust AVM of one-step estimates (x100000) CONST prorat lnmu 724.952 -116.944 34.748 -786.171 104.774 1550.615 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++t+++++ TABLE 19 ELECTRONICS SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 73 Sample period is 1988 to 1990 Observations: 163 Degrees of freedom: 158 Dependent variable is: dloginv Instruments used are: CONST proratla lnmuls TIM DUMS ------ONE-STEP ESTIMATES RSS = 71. 462719 TSS = 73.795977 Estimated sig ma-squared (levels) = 0.226148 Wald test of joint si gnificance; 2.953471 df =2 Wald test - it sig of time dums: 5.042572 df =3 Var Coef Std. Error T-Stat P-Value CONST 0.228658 0.108565 2.106194 0.035187 prorat 0.014714 0.008952 1.643639 0.100251 lnmu -0.001651 0.109294 -0.015105 0.987949 D89 -0.178467 0.158212 -1.128027 0.259309 D90 -0.219552 0.128243 -1.712004 0.086896 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -3.455 C 52 3 Test for second-order serial correlation: 1.710 [ 38 3 ------ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 0.841852 df = 2 Wald test - it sig of time dums: 5.761961 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.228658 0.110555 2.068272 0.030614 prorat 0.014714 0.017835 0.825006 0.409368 Inmu -0.001651 0.127076 -0.012991 0.989635 D89 -0.178467 0.143222 -1.246084 0.212734 D90 -0.219552 C). 108394 -2.025493 0.042817 Robust t est for first-order serial correlation: -1.914 [ 52 3 Robust t est for second-order serial correlation: 1.717 [ 38 ] Estimated serial correlation matrix 1.000 1.000 -0.222 0.257 - 0.492 1.000 Plumber of observations available to sample covariances 50 42 52 38 48 61 Model just identified - two-step estimates and one-step estimates coincide ------TASLE 20 ELECTRONICS SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv C). 12571 0.67493 -2.26800 2.44715 prorat 2.50357 4.82378 -1.41262 46.03984 1n mu 0.41392 0.42540 0.00000 1.91044 Correlation Matrix dloginv prorat Inmu 1.00 0.11 1.00 -0.01 -0.19 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST prorat lnmu D89 D90 1178.629 -29.572 8.014 1194.520 -531.433 27.751 2503.090 -1261.127 -1.080 -10.130 -682.801 -0.494 -45.741 925.826 1644.618 Robust AVM of one- step esti mates (x1 00000) CONST prorat lnmu D89 D90 1222.247 -108.698 31.808 1614.837 -755.730 96.284 11.484 -945.623 20.590 2051.260 -545.047 -11.815 -59.876 413.112 1174.934 t+t+t+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 21 CLOTHING SECTOR D. P. D. RESULTS LEVELS IV Number of firms: 53 Sample period is 1988 to 1990 Observations: 123 Degrees of freedom: 119 Dependent variable is: loginv Instruments used are: CONST lnkhlag proratla lnmula ------ ONE-STEP ESTIMATES 115.218683 TSS 142 RSS = = . 480903 (levels) Estimated sigma-squared =0 . 968224 Wald test of joint significance: 52.000751 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.184326 1.163379 0.158440 0.874110 laglogin 0.985627 0.145428 6.777400 0.000000 prorat 0.049803 0.026546 1.876129 0.060638 lnmu 0.054561 0.305055 0.178856 0.858051 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -2.213 C 41 ] Test for second-order serial correlation: -1.770 [ 29 3 ------ ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 43.785445 df 3 Var Coef Std. Error T-Stat P-Value CONST 0.184326 1.297000 0.142117 0.886987 laglogin 0.985627 0.167070 5.899492 0.000000 prorat 0.049803 0.035182 1.415579 0.156899 lnmu 0.054561 0.237616 0.229617 0.818389 Robust test for first-order serial correlation: -2.026 C 41 3 Robust test for second-order serial correlations -1.491 C 29 3 Estimated serial correlation matrix 1.000 1.000 -0.277 1.000 -0.405 -0.382 Number of observations available to sample covariances ,39 Z4 41 29 36 43 Model just identified - two-step estimates and one-step estimates coincide ------TABLE 22 CLOTHING SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max loginv 7.65923 1.08068 5.40268,10.71250 laglogin 7.42861 1.20292 2.56495 10.71250 prorat 2.64756 5.10037 -0.98204 35.68221 lnmu 0.38876 0.37268 0.00000 1.54872 Correlation Matrix loginv laglogin prorat lnmu 1.00 0.61 1.00 -0.02 -0.24 1.00 -0.02 0.05 -0.27 1.00 ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu j35345.176 -16738.811 2114.944 -1895.034 212.774 70.467 -13362.277 1194.543 328.883 9305.879 Robust AVM of one-step estim ates (x100000) CONST laglogin prorat lnmu 168220.923 -21481.590 2791.231 -3027.811 347.631 123.779 -4821.920 216.211 390.165 5646.156 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 23 CLOTHING SECTOR D. P. D. RESULTS LEVELS IV Number of firms: 53 Sample period is 1988 to 1990 Observations: 123 Degrees of freedom: 117 Dependent variable is: loginv Instruments used are: CONST lnkhlag proratla lnmula TIM DUMS ------ONE-STEP ESTIMATES RSS = 113.975244 TSS = 142.480903 Estimated sigma-squared (levels) 0.974147 Wald test of joint significance: 49.259641 df = 3 Wald test - it sig of time dums: 7.130057 df = 2 Var Coef Std. Error T-Stat P-Value CONST 0.107130 1.194270 0.089703 0.928523 laglogin 1.032241 0.156853 6.580951 0.000000 0.056679 0.027194 prorat 2.084238 " 0.037139 lnmu 0.107080 0.311662 0.343577 0.731164 D89 -0.633829 0.242712 -2.611448 0.009016 D90 -0.275820 0.242481 -1.137491 0.255333 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -1.815 C 41 3 Test for second-order serial correlation: -1.854 C 29 3 ------ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 37.181121 df = 3 Wald test - it sig of time dums: 6.854276 df = 2 Var Coef Std. Error T-Stat P-Value CONST 0.107130 1.363247 0.078584 0.937363 laglogin 1.032241 0.188055 5.489028 0.000000 prorat 0.056679 0.036498 1.552927 0.120441 lnmu 0.107080 0.241017 0.444283 0.656838 D89 -0.633829 0.275307 -2.302265 0.021320 D90 -0.275820 0.281786 -0.978829 0.327664 Robust test for first-order serial correlation: -1.458 C 41 ] Robust test for second-order serial correlation: -1.512 C 29 ] Estimated serial correlation matrix 1.000 -0.170 1.000 1.000 -0.412 -0.391 Number of observations available to sample covariances 39 34 41 29 36 43 two-step Model just identified - estimates and one-step estimates coincide ------TASLE 24 CLOTHING SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max loginv 7.65923 1.08068 5.40268 10.71250 laglogin 7.42861 1.20292 2.56495 10.71250 prorat 2.64756 5.10037 -0.98204 35.68221 inmu 0.38876 0.37268 0.00000 1.54872 Correlation Matrix loginv laglogin prorat lnmu 1.00 0.61 1.00 -- , -0.02 -0.24 1.00 -0.02 0.05 -0.27 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu D89 D90 142628.019 -18446.934 2460.280 -2063.751 246.829 73.953 -15132.398 1543.899 364.137 9713.319 9327.897 -1552.567 -156.218 -1517.842 5890.888 9829.961 -1618.166 -157.702 -1633.009 3562.866 5879.716 Robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu D89 D90 185844.190 -25289.736 3536.478 -3446.292 430.535 133.214 -8153.932 761.835 475.023 5808.941 19079.191 -3003.345 -434.152 -2275.677 7579.367 20136.788 -3261.441 -335.687 -968.921 5976.377 7940.323 TT T TtTT tt t tt 11, It, tt 1- 1- tttttt tt-I"}'I- ...... "..... "".... "rsrTTTTTTTTTT TT -V -r TASLE 25 CLOTHING SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 53 Sample period is 1988 to 1990 Observationss 123 Degrees of freedom: 119 Dependent variable is: dloginv Instruments used are: CONST lnkhlag proratla lnmula ------ ONE-STEP ESTIMATES TSS 125 RSS = 115.218683 = . 947474 Estimated sigma-squared (levels) =-0 . 484112 Wald test of joint significance: 12.739973 df =3 Var Coef Std. Error- T-Stat P-Value CONST 0.184326 0.928495 0.198522 0.842637 laglogin -0.014373 0.114487 -0.125544 0.900093 0.019788 0.011843 prorat 0.049803 2.516802 lnmu 0.054561 0.220024 0.247977 0.804152 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -2.103 C 41 ] Test for second-order serial correlation: -1.762 E 29 ] ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 41.830350 df =3 Var Coef Std. Error T-Stat P-Value CONST 4.184326 0.814687 0.226254 0.821004 0.103902 0.889976 l ag l og in -0.014373 -0.138334 0.049803 0.011571 4.304151 0.000017 prorat . lnmu 0.054561 0.153087 0.356405 0.721538 Robust test for first-order serial correlation: -1.778 C 41 ] Robust test for second-order serial correlation: -1.407 C 29 3 Estimated serial correlation matrix i. 000 1.000 -Q . 277 1.000 -0.445 -0.382 Number of observations available to sample covariances Z9 34 41 29 36 43 Model just identified - two-step estimates and one-step estimates coincide ------TABLE 26 CLOTHING SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv 0.23062 1.01605 -2.04157 4.07362 laglogin 7.42861 1.20292 2.56495 10.71250 prost 2.64756 5.10037 -0.98204 35.68221 lnmu 0.38876 0.37268 0.00000 1.54872 Correlation Matrix dloginv laglogin prorat lnmu 1.00 -0.54 1.00 0.27 -0.24 1.00 -0.09 0.05 -0.27 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu 86210.325 -10548.960 1310.729 -1303.348 150.095 39.158 -9249.823 916.771 200.839 4841.057 Robust AVM of one-step estimates (xiOC)000) CONST l aglogin prorat inmu 66771.430 -8412.892 1079.554 -698.506 84.619 13.389 -2265.295 165.849 55.635 2343.560 TT 41 . +.... + FtFF I. ifi .. fiit. 4 i+ FFF.. +.. F... hf+. ++ F. FhF t 'T 1' T' TTTTTTTT1TTTTt? TABLE 27 CLOTHINS SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 53 Sample period is 1988 to 1990 Observations: 123 Degrees of freedom: 117 Dependent variable is: dloginv Instruments used ere: CONST lnkhlag proratla lnmula TIM DOMS ------ONE-STEP ESTIMATES RSS = 113.975244 TSS = 125.947474 Estimated sigma-squared (levels). = 0.487074 Wald test of joint si gnificance: 13.491017 df = 3 Wald test - it sig of time dums: 5.503526 df = 3 Var Coef Std. Error- T-Stat P-Value CONST 0.107130 0.946258 0.113214 0.909861 laglogin 0.032241 0.122079 0.264096 0.791706 0.020402 0.005468 prorat 0.056679 2.778069 inmu 0.107080 0.226397 0.472974 0.636232 D89 -0.633829 0.277427 -2.284670 0.022332 D90 -0.275820 0.232893 -1.184322 0.236286 to NOTE: Standard errors and test statistics not robust heteroskedasticity Test for first-order serial correlation: -1.761 C 41 Test for second-order serial correlation: -i. e6o C 29 3 ------ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 57.487634 df = 3 Wald test - it sig of time dums: 7.239915 df =3 Var Coef Std. Error T-Stat P-Value CONST 0.107130 0.948545 0.112941 0.910077 laglogin 0.032241 0.129759 0.248465 0.803775 prorat 0.056679 0.013807 4.105135 0.000040 inmu 0.107080 0.177697 0.602599 0.546775 D89 -0.633829 0.274422 -2.309683 0.020906 D90 -0.275820 0.287353 -0.959867 0.337122 Robust test for first-order serial correlation: -1.334 C 41 ] Robust test for second-order serial correlation: -1.453 C 29 ] Estimated serial correlation matrix 1.000 -0.170 1.000 1.000 -0.412 -0.7,91 Number of observations available to sample covariances :39 34 41 29 36 43 Model just identified - two-step estimates and one-step estimates coincide ------TABLE 28 CLOTHING SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Ma>; dloginv 0.23062 1.01605 -2.04157 4.07362 laglogin 7.42861 1.20292 2.56495 10.71250 prorat 2.64756 5.10037 -0.98204 35.68221 lnmu 0.38876 0.37268 0.00000 1.54872 Correlation Matrix dloginv laglogin prorat lnmu 1.00 -0.54 1.00 0.27 -0.24 1.00 -0.09 0.05 -0.27 1.00 ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST laglogin prorat lnmu D89 D90 89540.345 -11343.550 1490.326 41.626 -1394.370 171.814 225.571 5125.580 -10198.819 1134.684 4555.634 -1041.267 -133.167 -1195.355 7696.559 4898.542 -947.547 -112.507 -1261.218 3245.061 5423.909 Robust AVM of one-step estimates (x100000) CONST laglogin prorat Inmu D89 D90 89973.781 1683.742 -12069.998 150.670 19.063 -1115.750 106.594 3157.621 -5296.271 650.077 10038.545 -1862.169 -168.293 -1470.038 7530.764 8642.627 -1731.074 -158.446 -1299.568 5528.482 8257.151 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 29 CLOTHING SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 53 Sample period is 1988 to 1990 123 Degrees freedom: Observations: of _120 Dependent variable is: dloginv Instruments used are: CONST proratla lnmula ------ ONE-STEP ESTIMATES RSS = 117.227401 TSS = 125.947474 Estimated sigma-squared (levels) = 0.488448 Wald test of joint significance: 9.149326 df =2 Var Coef Std. Error T-Stat P-Value CONST 0.074526 0.124237 0.599873 0.548591 prorat 0.050326 0.017389 2.894093 0.003803. lnmu 0.058787 0.211064 0.278527 0.780608 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -2.121 C 41 ] Test for second-order serial correlation: -1.745 C 29 3 ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 30.392975 df =2 Var Coef Std. Error T-Stat P-Value CONST 0.074526 0.092239 0.807970 0.419108 prorat 0.050326 0.009302 5.409971 0.000000 lnmu 0.058787 0.157939 0.372214 0.709734 Robust test for first-order serial correlation: -1.923 C 41 ] Robust test for second-order serial correlation: -1.387 C 29 3 Estimated serial correlation matrix 1.000 -0.279 1.000 1.000 -0.402 -0.376 Plumber of observations available to sample covariances 39 Z4 41 29 Z,6 43 two-step Model just identified - estimates and one-step estimates coincide ------TAOLE 30 CLOTHING SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv 0.23062 1.01605 -2.04157 4.07362 prorat 2.64756 5.10037 -0.98204 35.68221 lnmu 0.38876 0.37268 0.00000 1.54872 Correlation Matrix dloginv prorat lnmu 1.00 0.27 1.00 -0.09 -0.27 1.00 ------ ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100000) CONST prorat lnmu 1543.473 -138.463 30.239 -2107.492 138.602 4454.795 Robust AVM of one-step estimates (x100000) CONST prorat lnmu 850.798 -46.845 8.654 -986.191 37.939 2494.469 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ TABLE 31 CLOTHING SECTOR D. P. D. RESULTS FIRST DIFFERENCES IV Number of firms: 53 Sample period is 1988 to 1990 Observations: 123 Degrees of freedom: -118 Dependent variable is: dloginv Instruments used are: CONST proratla lnmula TIM DUMS ------ONE-STEP ESTIMATES RSS = 109.693020 TSS = 125.947474 Estimated sigma-squared (levels) =-0.464801 Wald test of joint significance: 11.396426 df = 2 Wald test - it sig of time dums: 5.784191 df = 3 Var 'Coef Std. Error T-Stat P-Value CONST 0.340421 0.180379 1.887252 0.059126 prorat 0.055356 0.016929 3.269843 0.001076 lnmu 0.095473 0.207265 0.460631 0.645063 D89 -0.613535 0.258015 -2.377903 0.017411 D90 -0.254475 0.214522 -1.186241 0.235527 NOTE: Standard errors and test statistics not robust to heteroskedasticity Test for first-order serial correlation: -1.823 C 41 ] Test for second-order serial correlation: -1.920 [ 29 ] ------ONE-STEP ESTIMATES WITH ROBUST TEST STATISTICS Wald test of joint significance: 32.403434 df = 2 Wald test - it sig of time dums: 7.494676 df = 3 Var Coef Std. Error T-Stat P-Value CONST 0.340421 0.182675 1.863532 0.062387 prorat 0.055356 0.009921 5.579749 0.000000 lnmu 0.095473 0.159165 0.599836 0.548616 D89 -0.613535 0.237201 -2.586558 0.009694 D90 -0.254475 0.254457 -1.000071 0.317276 Robust test for first-order serial correlation: -1.435 C 41 3 Robust test for second-order serial correlation: -1.471 C 29 ] Estimated serial correlation matrix 1.000 -0.170 1.000 1.000 -0.423 -0.405 Number of observations available to sample covariances 39 3,4 41 29 36 43 Model just identified - two-step estimates and one-step est imates coincide ------TAELE 32 CLOTHING SECTOR DESCRIPTIVE STATISTICS Variable Mean Std Dev Min Max dloginv 0.23062 1.01605 -2.04157 4.07362 prorat 2.64756 5.10037 -0.98204 35.68221 lnmu 0.38876 0.37268 0.00000 1.54872 Correlation Matrix dloginv arorat lnmu 1.00 0.27 1.00 -0.09 -0.27 1.00 ASYMPTOTIC VARIANCE MATRICES Non-robust AVM of one-step estimates (x100040) CONST prorat lnmu D89 D90 3253.665 -122.305 28.660 -1705.112 132.803 4295.874 -3178.852 -19.834 -423.680 6657.182 -2221.899 -0.345 -500.110 2462.323 4601.967 Robust AVM of one-step estimates (x100000) CONST prorat lnmu D89 D90 3337.020 -29.619 9.842 -590.283 47.395 2533.342 -3254.020 -38.155 -666.936 5626.443 -3665.682 -34.560 -610.556 3742.469 6474.824 175 APPENDIX 2 The data set has been obtained from the volumes of years 1988, 1989,1990,1991 of Mediobanca survey on the accounting data of the main Italian enterprises. This choice is due to the fact that since 1988 the lower bond of the sample has been raised to 25 billion Italian lire. Since each volume contains the data of the two previous years, the observations refer to years 1986,1987,1988, 1989,1990, i. e. 5 years. Since the variable I(t) has to be constructed with the stock values of K (capital) and "accumulated depreciations" for t and t-1, plus the flow variable "depreciations" at time "t", the number of available years for the estimates would reduce to 4, but since the empirical specification employed here contains the lagged dependent variable, then the number of years available is 3. Furthermore, given that the dependent variable has to be constructed by using the values of consecutive stock values, and given that this same variable appears (lagged) as a regressor, then for each firm to be included in the dataset, it is necessary to have at least three data referred to three consecutive years. Therefore the three industrial samples (chemical, electronics and clothing sectors) have been, constructed by considering all of the firms appearing in the for Mediobanca survey, which at least three years of consecutive (between 1986 observations and 1990) were available. For this purpose an unbalanced panel data sample has been created. Unlike the includes balanced sample, which only the "survivors" over the sample period, and, for this reason, might contain some bias and lose the firms information referred to the that exit' an to the new entrants 176 during the time under consideration, the unbalanced sample yields more complete information, although at the cost of using a more complex econometric methodology. In any case, some limitations have been necessary in order to overcome a few problems that have arisen. First of all, we have to keep in mind that the purpose of this paper is to analyze the investment in physical capital. Therefore investment any kind of financial has not been taken into account for the estimates of the investment function. For- the same reason, it has been necessary to exclude from the dataset those firms that, during the period under consideration, have modified their nature from industrial firms to financial holdings. In any case, the information relative to their behaviour has been kept, by including in the samples (when it was possible) all of the fitms belonging to the old and new born financial holdings and operating in the industries under consideration. Obviously, each individual firm has to keep its individual nature over the time under consideration, and for this reason, mergers have been regarded as events that modify the individual nature of each firm. The unbalanced. panel technique individuals allows in fact to consider as separate the firms before the mergers and the new-born firms that are determined by mergers. Such an approach has also some common sense validity, to the extent determined by that the individual the merger is actually different - in its in its behaviour and conjectures - from the different individuals that contribute to determine the merger. Furthermore, a in merger introduces, general, an unpredictable piece of information in the information set of the different agents. In each different the year under consideration conjectures might be modified, and this 177 is consistent with the argument that justifies the "partial adjustment" empirical specification employed in this paper for the investment function. A last point to mention is the fact that the process of liquidation that precedes mergers or failures might actually start before the merger and/or the failure of the firm under consideration takes place in legal terms. In other words, the data of the firms under consideration showed, in the years just before the liquidation or the merger, the typical aspects that characterize the processes of liquidation or merger. Such processes (which in any case refer to events reported in the original volumes of Mediobanca survey) typically involve drastic changes in the balance sheet structure of the firm, like a dramatic increase in the "financial assets" or in the "other, assets" associated to a very large reduction in the physical capital. Obviously for each unity, of observation, the years where such phenomena took place could not be taken into account, while the unity of observation itself could still be considered for the rest of the period, where no mergers or liquidations took place. However, in all of the three samples, the number of firms involved in phenomena of mergers and/or liquidation is very small and neglectable, compared to size of the samplel3. The mere changes of name or denomination that do not affect the 13 In'any case, the raw data (unfortunately not available on diskette) of the original volumes of Mediobanca survey on the Italian firms with more than 25 billion of lira sales may be obtained by contacting the research office of Mediobanca at the following address: Ufficio Studi di Mediobanca via Filodrammatici 20100 Milano Italy - tel. (+)39 (0)2 88291 Alternatively, the same volumes may be obtained by the author of this thesis. 178 structure, the main business and the characteristics of the individual firms will obviously not be regarded as mergers. The industry classification in the Mediobanca sample does not always correspond to the standard NACE, although this should not constitute a problem for the chemical sector. In order to simplify the exposition of the dataset, the list of the firms included in the sample is reported separately; each firm is associated to a number, that will be used to identify the data in the tables of the dataset. Since the econometric analysis has been data firm (i. firm performed with panel techniques, each e. each (corresponding line in the data number) refers to several years to a tables), and each year contain its relative values for the different follows the firms included in the industry variables. In what listed to their samples are and associated reference number tables the data. appearing in the of Chemical Sector: 1 Procter & Gamble Italia; 2 Exxon Chemical Mediterranea; 3 Henkel Sud; 4 Agfa Gevaert; 5 Alusuisse Italia; 6 Colgate-Palmolive; 7 Industrie Vernici Italiane; 8 Chimet; 9 Hoechst Italia; 10 Elettrocarbonium; 11 SIAPA - Italo Americana Prodotti Antiparassitari; 12 Henkel Chimica; 13 Italiana Coke; 14 SPAD - Piemontese Amidi e Derivati; 15 Caffaro; 16 ACNA Chimica Organica; 17 Snia Tecnopolimeri; 18 Grace Italiana; 19 Johnson Wax; 20 Sorin 22 Biomedica; 21 Maxmeyer- Duco; Annunziata; 23 General Electric 24 plastics-Italia; Monsanto Italiana; 25 Istituto delle Vitamine; 26 Abet Laminati; 27 Marchon Italiana; 28 Rivoira; 29 Dobfar; 30 Comind Sud; 31 Liri Industriale; 32 Zanussi Componenti Plastica; 33 34 Uniroyal Chimica; Giovanni Crespi; 35 Tillmanns; 36 38 Italesplosivi; 37 Vitrofil; Oxon Italia; 39 Salchi; 40 Engelhard; 41 Baslini Industrie Chimiche; 42 PCBI; 43 Carbochimica Italiana; 44 SIPE Nobel - Italiana Prodotti Esplodenti; 45 Degussa Prodotti Ceramici; 46 Mapei; 47 Hoechst Sara; 48 Marchon Sud; 49 Novacrome; 50 Mas Industriale; 51 Laminati Plastici e Rivestimenti; 52 Baldini; 53 Ilford Photo; 54 FIAP - Fabbrica Italiana Articoli Plastici; 55 vifan; 56 Simel; 57 Diversey; 58 Vernici Lalac; 59 Ecofuel; 60 sun Chemical Inchiostri (called Baglini Inchiostri until 1988); 61 ICI (from Italia; 62 Enichem Augusta 1990 called Enimonst Augusta); 63 65 Kodak; 64 Enichem Tecnoresine; Sandoz Prodotti Chimici; 66 SIAC - 179 Societä Italiana Additivi per Carburanti; 67 Terni Industrie Chimiche; 68 COMET S. A. R. A.; 69 Union Carbide Italia; 70 Sikkens Linvea; 71 Miles Italiana; 72 Veneziani; 73 Autoadesivitalia; 74 Sarma; 75 Alta; 76 Atochem Italia; 77 BASF Vernici; 78 Paraffine Sarde; 79 Nuova Pansac; 80 Boston; 81 FAR - Fabbrica Adesivi Resine; 82 SIAD - Italiana Acetilene e Derivati; 83 Brill; 84 Henkel Italiana; 85 Fotoindustria; 86 Gorlex; 87 Flexa Films; 88 Romana Chimici; 89 Vedril; 90 Dow Italia; 91 SIO - Industria Ossigeno Altri Gas; 92 Mazzucchelli Celluloide; 93 Bristol Europe; 94 Italiana Keller; 95 LATI - Industria Termoplastici; 96 Orsa; 97 NALCO Italiana; 98 Boero Colors; 99 Silo; 100 3M Italia; 101 Agrimont; 102 Erba Biochimica; 103 Fanini Fain - Fabbrica Italiana Articoli Novitä; 104 Luigi Stoppani; 105 AKZO Chemicals; 106 Cartochimica Valpellice; 107 IMEXCO Specialties; 108 BASF Italia; 109 Enichem Synthesis; 110 CERESTAR Italia; 111 DSM Italia; 112 Manitoba Italia; 113 Samatec; 114 Casco Nobel; 115 Industrie generali; 116 Seeber; 117 Roussel Hoechst Agrovet; 118 Alfatherm Industriale; 119 Giuseppe Olmo Superflexite Italiana; 120 Ovatex; 121 ITB; 122 Crion; 123 Zobele Industrie Chimiche; 124 Flexa. Electronics Sector: 1 ITALTEL SIT; 2 BULL HN Information System Italia (called until 1987 Honeywell Bull Italia); 3 ELSAG Elettronica San Giorgio; 4 Digital Equipment; 5 Robert Bosch; 6 Siemens Telecomunicazioni; 7 Texas Instruments Italia; 8 Italtel Telematica; 9 UNISYS Italia; 10 SOGEI; 11 Nuova Magrini Galileo; 12 Siemens Data; 13 Contraves Italiana; 14 ITALSIEL - Italiana Sistemi Informativi Elettronici; 15 Elettronica; 16 Canon Italia; 17 Selenia Spazio (from 1990 Alenia Spazio); 18 Telemecanique; 19 Arcotronics Italia; 20 Hantarex; 21 DEA Digital Electronics Automation; 22 Honeywell; 23 Marposs; 24 SIES Peterlongo; 25 Marelli Autronica; 26 Italtel Tecnoelettronica; 27 CSELT Centro Studi e Laboratori Telecomunicazioni; 28 Dataconsyst; 29 Onceas; 30 Teknecomp; 31 Italdata; 32 Apple Computer; 33 SAFNAT - Fabbrica Nazionale Apparecchi Telefonici; 34 35 Data Management; Elettronica Industriale; 36 ARE - Applicazioni Radio Elettroniche; 37 Hartmann & Braun Italia; 38 OSAI A-B; 39 IBM Italia; 40 Siemens; 41 Selenia - Industrie Elettroniche Associate; 42 ASEM; 43 Industrie Magneti Marelli; 44 Marconi Italiana; 45 Hitachi Sales Italiana; 46 Plessey; 47 SMA Segnalamento Marino ed Aereo; 48 Compugraphics Italia (from 1989 Agfa Compugraphic); 49 Varian; 50 Procond Elettronica; 51 GTE Sylvania; 52 Fabbrica Accumulators York; 53 Industrie Face Standard; 54 Hewlett-Packard italiana; 55 Landis & Gyr; 56 OLTECO - Olivetti Telecomunicazioni; 57 Bonfiglioli Riduttori; 58 Dial Telecomunicazioni; 59 Fracarraro Radioindustrie; 60 Nuova Industrie Elettriche di Legnano; 61 Telettra; 62 SGS Thomson Microelectronics; 63 Sony Italia; 64 Olivetti Canon Industria; 65 Elmer; 66 Bailey Esacontro; 67 Esaote 68 69 Nashua Biomedica; AROS; Reprographics; 70 Necsy - Network Control Sytem; 71 SEPA - Societä di Elettronica per 1'Automazione; 72 Philips Automation; 73 Olivetti Prodotti Industriali. 180 Clothing Sector: 1 Benetton; 2 GFT; 3 Stefanel; 4 Levi Strauss Italia; 5 Prenatal; 6 Maglificio Bellia; 7 Maglificio Calzificio Torinese; 8 Byblos; 9 Sanremo Moda Uomo; 10 Golden Lady; 11 Fila Sport; 12 Fratelli Claudio e Carlalberto Corneliani; 13 Belfe; 14 Giole; 15 Forall Confezioni; 16 Calze Malerba; 17 SICEM; 18 Dalmas; 19 IME - Industria Maglieria Europea; 20 Cagi Maglierie; 21 Sandys; 22 Lubiam Moda per 1'Uomo; 23 Calzificio di Parabiago Mario Re Depaolini; 24 Belvest; 25 Confezioni F. G.; 26 GM; 27 Sima; 28 Industria Adriatica Confezioni; 29 Facib di Cortesi & C.; 30 Max Mara; 31 Carrera; 32 Marina Rinaldi; 33 Manifatture del Nord; 34 La Matta; 35 Genny Moda; 36 CP Company; 37 Samar; 38 Maska; 39 Columbia; 40 Confar Confezioni Aretine; 41 La Granda Confezioni; 42 Incom - Industria Confezioni Montecatini; 43 Commerciale Abbigliamento; 44 Maglieria ragno; 45 Ligron; 46 Ball; 47 GFT Donna; 48 Luisa Spagnoli; 49 Lovable Italiana; 50 Filodoro Calze; 51 Luck; 52 Hitman Industria Confezioni; 53 Calzificio Fratelli Carabelli. Chemical sector firm t-1 D. DEPR. K(-1) no. year V ACC. ACC. D. INV. V. ADD. LAD. C. VAR. PR. i' 1987 163397 127710 32908 151106 99852 17341 140297 62090 78207 1988 175998 14 314 24143 163397 127710 22140 121024 70803 50221 1989 195694 155397 2 748 175998 142314 29361 171466 86963 84523 1990 300011 196865 30015 195694 155397 9 864 201923 102673 99250 '2 1987 43596 30188 4555 38458 28659 8164 29221 22190 7031 1988 58393 31390 3585 43596 30188 17180 36047 26204 9843 1989 69223 33009 4373 58393 31390 13584 39984 29369 10615 1990 76549 34885 4965 69223 33009 10415 41204 31669 9535 3 1987 68393 40684 6499 64059 36061 6210 50025 26159 23866 1988 74576 45563 6420 68393 40684 7724 33730 27851 5879 1989 81271 48292 3966 74576 45563 7932 25873 30010 -4137 1990 86624 55614 6903 81271 48292 4934 45440 32516 12924 4 1987 18381 10029 1651 17487 8996 1512 58103 29715 28388 1988 19583 10342 1501 18381 10029 2390 61566 31435 30131 1989 20980 10752 1234 19583 10342 2221 63874 35407 28467 1990 24898 116.31 2359 20980 10752 5398 71180 38324 32856 176199 139468 19441 155631 118668 19209 75767 36265 39502 5 1987 22261 1988 205470 161217 176199 139468 29783 94486 38316 56170 1989 227824 180479 19793 205470 161217 22885 87736 43869 43867 1990 250930 194082 14203 227824 180479 23706 71958 47489 24469 6 1987 77376 36271 3025 68693 33685 9122 63334 45759 17575 1988 86041 40534 4497 77376 36271 8899 61620 51052 10568 1989 92792 46732 6946 86041 40534 7499 76563 55021 21542 1990 99902 54083 7749 92792 46732 7508 85889 57999 27890 7 1987 224976 130383 18402 210878 113526 15643 84161 56264 27897 1988 226752 142774 19167 224976 130383 8552 98191 60285 37906 1989 237936 157501 16893 226752 142774 13350 111122 64734 46388 1990 263644 173389 17479 237936 157501 27299 123419 68730 54689 8 1987 18724 10776 1889 14980 9495 4352 9387 3379 6008 1988 21133 12810 2264 18724 10776 2639 11962 3803 8159 1989 23009 14546 2236 21133 12810 2376 13145 4422 8723 1990 24843 16648 2052 23009 14546 1784 11148 5011 6137 9 1987 48459 33131 3452 51335 35152 2597 74133 59643 14490 1988 49718 33104 2967 48459 33131 4253 75386 65380 10006 1989 50141. 32200 3371 49718 33104 4698 78077 74172 3905 1990 43346 25729 3735 50141 32200 3411 70791 79353 -8562 io 1987 229072 174233 15099 218982 162650 13606 77288 56843 20445 1988 239503 189978 16655 229072 174233 11341 82460 53488 28972 1989 253467 201898 12551 239503 189978 14595 101403 60041 41362 1990 270382 213614 12756 253467 201896 17955 84636 65079 19557 38998 3242B 2649 34473 20273 5019 53229 34681 18548 il 1987 . 1988 43230 34692 2550 38998 32428 4518 54162 38128 16034 2817 43230 61517 1989 44700 35929. 34692 3050 44997 16520 1990 5026 36731 3422 44700 35929 8146 65149 48384 16765 5993 55527 12 1987 66232 26101 20624 11221 35297 20179 15118 6908 66232 1988 79111 3227B . 2610i 13610 39887 26127 13760 1989 94468 39209 7994 '79111 32278 16420 42250 29576 12674 1990 107540 47301 9027 94468 39209 14007 47810 31463 16347 9512 132423 i., 1987 143926 110383 100614 11246 58243 49004: ' 9239 1988 149177 118878 9334 143926 1107,83 5090 60291 48013 12 278 1989 144737 120124 7639 148177 118878 2953 63876 48513 15363 1990 129740 106916 6187 144737 120124 4398 55658 42729 12929 14 1987 85844 67436 10647 77315 59954 11694 31738 21375 10363 1988 99152 75684 10236 85844 67436 15296 48326 23155 25171 1989 112333 83198 11285 99152 75684 16952 61655 25422 36233 1990 126765 92024 11732 112333 83198 17338 60431 27509 32922 15 1987 125042 99573 6971 119552 93596 6484 67983 33169 34814 1988 130867 107407 9820 125042 99573 7811 69132 34708 34424 1989 138528 116328 10608 130867 107407 9348 715.5 36042 35493 1990 145610 126691 11490 138528 116328 8209 79069 40323 38746 177690 62325 21357 156861 42303 22164 66207 35875 30332 16 1987 62325 1988 217191 75471 15473 177690 41828 45593 37698 7895 1989 271008 91252 18751 217191 75471 56787 4025 37761 -33736 1990 316992 109657 21024 271008 91252 48603 2966 42274 -39308 17 1987 83380 36708 9810 71353 28331 13460 48104 19747 28357 1988 90072 45599 9322 83380 36708 7123 56463 22368 34095 1989 99668 55062 9837 90072 45599 9970 53826 25142 28684 1990 104426 65337 10744 99668 55062 5227 54026 27729 26297 18 1987 82100 27723 8783 75258 20074 7976 49515 35518 13997 1988 90374 33610 7420 82100 27723 9807 47952 40060 7892 1989 99009 40423 8060 90374 33610 9882 52812 45155 7657 1990 108886 51069 11408 99009 40423 10639 63539 47533 16006 19 1967 11688 5100 1584 10122 3914 1964 30169 13140 17029 1986 12784 6458 1709 11688 5160 1447 33771 13748 20023 1989 14281 7859 1448 12784 6458 1544 35799 15115 20684 1990. 15461 9138 1418 14281 7859 1319 40145 16633 23512 20 1987 53105 36903 9979 47400 29005 7786 65621 38284 27337 1988 52322 38878 8743 53105 36903 5985 80638 41360 39258 1969 64521 44972 7501 52322 38878 13606 77307 42734 34573 1990 80424 53033 9905 64521 44972 17747 86653 51670 34983 21 1987 51469 23357 5882 47891 18638 4741 43869 29390 14479 1988 56690 31138 5512 51469 23357 2952 50311 30646 19665 1989 64447 35426 5336 56690 31138 8805 53212 31428 21784 1990 71528 40112 6190 64447 35426 8585 50904 34153 16751 22 1987 54815 19792 2492 48037 17528 7006 26475. 14846 11629 1988 62380 21556 1928 54815 19792 7729 29133 15670 13463 1989 70153 23500 2490 62380 '21556 8319 36766 17869 18897 1990 83392 25788 2649 70153 23500 13600 37304 18983 18321 7475 4544 1347 7162 3238 354 10693 3903 6790 23 1987 , 1988 7662 4991 676 7475 4544 416 12473 4571 7902 1989 8259 5357 604 7662 4991 835 15524 5531 9993 1990 8662 5842 705 8259 5357 643 10689 6100 4589 24 1987 884 468 168 731 395 248 4311 3597 714 1988 1230 779 398 884 468 433 6009 4163 1846 1989 1524 947 255 1230 779 381 9331 5459 3872 1990 1700 1050 177 1524 947 250 9385 5791 3594 25 1987 13457 9246 1356 12490 8045 1122 13224 7798 5426 1988 14100 10218 1186 13457 9246 857 14318 8854 5464 1989 15779 11044 1131 14100 10218 1984 15156 9653 5503 1990 16356 11914 1158 15779 11044 865 15550 10908 4642 7826 30207 , 26 1987 46105 31688 24168 16204 45786 21674 24112 1988 54733 39509 9290 46105 31688 10097 52606. 24314 28292 1989 66026 47022 8858 54733 39509 12638 57814 27849 29965 1990 75127 53165 7659 66026 47022 10617 63190 31238 31952 55392 27 1987 59887 42768 5734 34540 2001 13893 7473 4 6420 1988 65340 50137 4861 59887 42768 2945 13245 7434 5811 1989 70011 54867 4894 65340 50137 4835 X5113 16537 8576 1990 72883 59321 4578 70011 54867 2996 24949 18559 6390 28 1987 57082 35615 5649 50990 29540 5666 38642 26637 12005 1988 64132 40203 5217 57082 35615 7679 35834 25525 10309 1989 71061 46027 7346 641: 2 4020,8451 37820 26201 11619 1990 82720 50964 750? 71061 46027 14225 38122 25757 12365 29 1987 8992 6467 2038 7207 4764 2120 1C077B 3460 7318 1988 11130 7890 2326 8992 6467 3043 11356 3032 8324 1989 13437 9754 2857 11130 7890 1300 9473 3459 6014 1990 17993 11645 2807 13437 9754 5472 6111 4082 2029 30-1987 73748 40554 11408 53594 29420 20428 23647 20103 3544 1988'115105 63830 15806 73748 40554 33887 53859 35179 18680 1989 167730 63247 21022 115105 63830 54230 81754 40103 41651 1990 204235 112609 29362 167730 83247 36505 76428 44309 32119 31 1987 39280 18350 3986 35335 14615 4196 20142 11173 8969 1988 44192 22377 4245 39280 18 354 5130 20000 11508 8492 1969 46104 26382 4657 44192 22377 2564 19509 10758 8751 1990 41578 24264 3289 46104, 26 382 881 14386 9460 4926 32 1987 24012 9256 2275 24040 8647 1638 25745 18627 7118 1988 25457 11087 2358 24012 9256 1972 26457 20196 6261 1989 29597 14032 3148 25457 11087 4343 28861 22260 6601 1990 34697 15572 2571 29597 14032 6131 26551 23494 3057 33 1987 26016 18653 2457 24382 16875 2313 17369 6184 11185 1988 28 355 21752 3283 26016 18653 2523 22827 9653 13174 1989 30340 24419 2706 28355 21752 2054 23521 10323 13198 1990 62306 31966 8198 30-340 24419 32617 23027 8139 14888 34 1987 43818 31561 4782 36439 26878 7478 28685 14805 13880 1.988 47657 35722 4751 43818 31561 4429 30379 15314 15065 1989 50810 39569 4331 47657 35722 3637 27839 16446 11393 1990 54205 42466 3666 50810 39569 4164 25541 17491 8050 35 1987 7950 5089 549 7757 4699 352 8364 4664 3700 1988 8172 5513 520 7950 5089 318 10593 5351 5242 1989 8462 5893 484 8172 5513 394 12071 6209 5862 1990 8596 6264 483 8462 5893 246 12692 6903 5789 36 1987 1732 791 195 1572 671 235 3033 2611 422 1988 1826 893 179 1732 791 171 2991 2742 249 1989 1853 1012 167 1826 893 75 2673 2491 182 1990 1859 1105 158 1853 1012 71 2871 2666 205 37 1987 43239 19143 4495 40920 15433 3104 29125 16277 12848 1988 63942 21981 3243 43239 19143 21108 31774 17295 14479 1989 65061 21413 7554 63942 21981 9241 37453 18983 18470 1990 67941 32888 11840 65061 21413 3245 43404 20253 23151 38 1987 83 375 56175 8133 75960 48884' 8253 24693 11357 13336 1988 88110 62811 7206 83375 56175 5305 21186 11402* 9784 1989 98883 72022 10014 88110 62811 11576 24806 12564 12242 1990 110042 78958 8559 98883 72022 12782 26852 14555 12297 39 1987 25013 14223 2480 12720 8507 9057 13606 9280 -4 26 1988 27998 17293 3316 25013 14223 3231 19340 13768 5572 1989 30401 20454 3433 27998 17293 2675 23112 15689 7423 3331 1990 34073 23503 30401 20454 3954 24178 17379 6799 40 1987 17113 9783 2387 15337 7487 1867 15134 10483 ' 4647 1988 18019 11717 2094 17113 9783 1066 16969 11778 5191 1989 19663 13830 2189 18019 11717 1720 18044 13066 4934 1990 20564 15381 1896 19663 13830 1246 19293 14592 4701 41 1987..,,, 16692- 10873 1376 16390 9574 379 11994 7926 --; 4064 1988 18801 12816 2024 16692 10873 2190 12348 7937 1989 21528 14415 1718 18801 12816 2846 13839 8622 1990 22326 14991 1772 21528 14415 1994 13282 E3246 42 1987 34515 30118 3538 32144 27469 3260 18627 5948 1988 40937 32681 3675 34515 30118 7534 24350 6751 32681 '19B9 527,33 36495 5166 40937 " 12748 21517 8029 1990 59874 42422 6889 52333 36495 8503 26508 8653 . 43 19(37 29755 18622 2063 28935 17139 1400 14352 7483 1966 31006 20088 1908 29755 18622 1693 16562 8420 1989 40345 22059 2366 31006 20088 9734 18636 8707 1990 45517 25488 3701 40345 22059 5444 16074 9607 44 1987 93995 59587 9185 89129 51703 6167 27135 14723 1986 97918 68505 9939 93995 59587 4944 28453 15320 1969 99365 74946 8911 97918 68505 3917 32567 16025 1990 96278 79939 8378 99365 74946 298 27569 16010 45 1987 29122 22396 1644 28300 20917 987 17895 11588 1988 30691 23732 1455 29122 22396 1688 18734 12955 1989 31029 24444 1477 30691 23732 1103 19506 14905 1990 33587 25170 1611 31029 24444 3443 20009 15358 46 1987 23285 14718 2713 18805 12128 4603 14192 6436 1988 24755 17105 2475 23285 14718 1558 13875 7784 2272 1989 26080 19033 24755 17105 1669 16704 9992 1990 28920 20629 2561 26080 19033 3805 23424 12370 47 1987 15930 10690 1972 14225 8723 1710 14192 5489 1988 17897 12611 1920 15930 10690 1966 13653 6162 1989 20091 14063 1707 17897 12611 2449 16195 7376 1990 26299 15776 1740 20091 14063 6235 18056 8426 48 1987 40356 27876 5734 37642 22163 2735 13893 7434 1988 42846 32731 4861 40356 27876 2496 13245 7473 1989 46331 37188 4457 42846 32731 3485 11432 7330 1990 52993 41109 3947 46331 37188 6688 13493 8075 49 1987 9020 6562 1723 8081 5503 1583 12371 6192 1988 10912 7705 1333 9020 6582 2102 12443 7034 1989 12272 8877 1274 10912 7705 1462 13774 8002 1990 14084 10134 1404 12272 8877 1959 13836 8847 50 1987 28999 18731 2368 25270 16909 4275 14612 10710 1988 37285 21541 3041 28999 18731 8517 14612 12010 1989 49725 22552 1491 37285 21541 12920 13571 11513 1990 50368 23591 1436 49725 22552 1040 6414 8339 _ 51 1987 23544 9232 3124 21211 6129 2354 10808 5794 1988 31743 12701 3509 23544 9232 8239 14595 6993 1989 36900 16658 4019 31743 12701 5219 16382 8298 1990 39905 20399 4498 36900 16658 3762 14246 8240 52 1987 11588 5421 817 10440 4743 1287 6458 4189 1988 12831 6635 1166 11588 5421 1195 7716 4674 1989 13751 7380 1075 12831 6635 1250 7732 5214" . 1990 14228 8574 1492 13751 7360 775 9108" 5641 53 1987 348 311 27 307 290 47 5124 2801 1988 451 327 26 348 '311 113 5071 3320 1989 601 426 101 451 327 152 6616 3836 1990 1071 607 191 601 -426 480 7318 4060 54 1967 18460 13884 1545 17169 12567 1519 10205 7460 1488 23239 15037 1016 18460 13884 4642 10087 8474 64129 16012 1091. 15037' 41006 12382 ',9850 1989 . -23239 1990 64940 18804 3121 64129 16012 1140 13043 10796 55 1987. 37884 15072 2787 17387 12352 20564 12343 4183 1988 4519= 18146 3089 37884 15072 7324 13989 5843 8146 1989 46414 169 363 5 45193 18146 1310 17262 6975 10287 1991 ß 469, .i. 532v. i 3830 46414 2169 708 12 638 7735 4903 56 1987 25653 4476 3272 23374 2009 3084 9030 5233 3797 1988 20586 7072 '3439 25654 4476 3776 9553 4578 4975 1989 33 432 10141 3911 28586 7072 5688 10184 6147 4037 3232 1990 36887 12514 33432 10141 4314 9057 6641 2416 57 1987 5342 3239 633 4958 2727 505 12719 9032 3687 1988 5644 3506 422 5342 3239 457 13677 9768 3909 . 1989 588 3546 375 5644 3,506 519 13361 100501 2060 199. ) 6435 3757 413 5828 3546 809 12938 11885 1053 58 1987 4303 3277 308 4092 2971 t:.213 7467 4330 3137 1988 5042 3665 479 4303 3277 830 8668 4922 3746 1989 5823 4175 617 5042 3665 888 8881 5253 628 1990 5858 4577 655 5823 4175 203 11093 6095 4998 59.1987 7789 1494 2567 8302 542 1102 3981 3857 124 1988 8162 2476 2752 7789 1494 2143 3703 5653 -1950 1989 8097 3494 2984 8162 2476 1901 14908 6346 8562 1990 9332 2115 3022 8097 3494 5636 12555 749. ^"_ 5063 23833 60 1987 25308 17661 2291 15978 2083 18030 11020 7010 1988 27246 18759 2002 25308 17661 2842 14574 9933 4641 1989 30543 20773, 2314 27246 18759 3597 14830 11281. 3549 1990 33393 23090 2618 30543 20773 3151 15418 12731 2687 61 1987 16467 7543 2025 12278 6246 4917 31782 26761 5021 1988 26255 9549 2905 16467 7543 10687 39264 36135 : 129 1989 52315 13541 5274 26255 9549 27342 51074 44871 6203 1990 59755 21394 9905 52315 13541 9492 63589 55650 7939 62 1987 192146 85505 23201 170960 64926 23808 87366 32714 54652 215666 103884 21459 192146 85505 26600 107720 : 8542 69178 1988 23 1989 251912 123406 L7 ý 215666 103884 39647 96623 39365 57258 63 1987 90121 50702 12441 82428 43808, 13240 112108 47631 64477 1988 97594 56402 11317 90121 50702 13090 108584 42022 66562 1989 110950 66886 14727 97594 56402 17599 107100 49605 57495 64 1987 95029 9788 14453 80585 2351 21460 23877 19163 4714 1988 116744 19733 16129 95029 9788 27899 25198 26445 -1247 1989 128985 27960 15671 116744 19733 19685 19129 29574 -10445 65 1987 33983 25260 1826 29898 19656 307 29480 20522 8958 1988 36118 26073 2027 33983 25260 3349 34132 22902 11230 1989 39168 27153 2690 36118 X607'_" 4660 38817 25889 12928 66 1987 23552 21966 1617 22401 20: 374 1176 31470 6448 25022 1988 24193- 23026 1060 23552 21966 641 27835 6910 20925 1989 0158 23830 806 24193 23026 967 27201 6686 20515 67 1987 186464 99120 8218 137587 92020 49995 33355 24277 9078 1988 217200 105612 11228 186464 99120 35472 42967 25074 17893 1989 240655 115784 12004 217200 105612 25287 51555 26890 24665 68 1987 35659 20425 3888 31698 15988 3412 18140 10618 7522 1988 39642 21607 2875 35659 20425 5676 15693 11331 4362 42112 23309 2879 39642 21607 3647 16013 11745 1989 . 4308 69 1987 63827 32471 2662 58955 30141 5204 25601 23744 1857 1988 145121 92573 10020 63827 32471 31212 54474 38676 15798 1989 150994 100227 9113 145121 92573 7332 56671 42601 14477 70 1987 10814 6333 1017 9900 5503 1101 20884 8120 12764 1988 11889 7230 963 10814 6333 1141 21991 9579 12412 1989 13808 8191 1160 11889 7230 2118 23088 10564 12524 71 1987 20963 14708 3830 18728 11142 2499 15048 7057 7991 22830 17382 3158 20963 14708 2351 14461 1988 , 8770 5691 .tom, "'-'- ý?:; a^ý 4Y,_ ý ;^, t .. X-- (ý . ýý_ tin.. aý, '. l :Ti !' .,. rr. ý, "a:. "a . 1989 24396 19103 2441 22830 17382 2286 16526 9885 6641 6485 72 1987 7572 4364 701 3756 1100 10534 6659 3875 1988 9086 4776 583 7572 4364 1685 8770 6986 1784 93 1989 10508 4335 743 9086 4776 2606 7448 2675 73 1987 18012 7964 1714 16336 6891 2317 8^9 5179 3113 1988 19500 9170 1797 18C012 7964 2079 8902 6162 2740 1989 22165 10256 1384 19500 9170 2963 8646 5836 2810 74 1987 11605 8493 1017 10093 7653 1689 9607 7489 2118 1988 12217 9585 1384 11605 8493 904 11696 7605 4091 1989 13822 10509 1399 12217 9585 2080 12811" 7845 4966 2089 21463 75 1987 23757 14480 12868 2771 7170 10444 -3274 1988 27245 16095 1741 23757 14480 3614 6099 1138"_ -5^84 1989 31443 16882 1411 27245 16095 4822 5630 9963 -. 4333 76 1988 616 402 80 572 326 48 10376 2951 7425 1989 604 425 80 616 402 45 7352 2911 4441 1990 477 327 81 604 425 52 7535 3354 4181 4796 32860 15402 77 1988 54082 28927 49658 25173 5466 48262 1989 61431 33657 5135 54082 28927 7754 51825 36960 14845 1990 68963 38718 5525 61431 33657 8016 55598 40114 15484 5742 39603 0 2257 34410 1047 33363 78 1988 40982 4864 1989 48412 9744 5763 40962 4064 8313 25887 1181 24706 1990 50501 15210 6350 48412 9744 2973 16266 1284 14982 2893 53773 21913 10372 79 1988 56254 15659 13655 3370 32285 1989 60648 18047 3398 56254 15659 5404 35180 22363 12817 1990 74855 27358 4680 60648 18047 9576 45311 29901 15410 3236 80 1988 45350 25132 38875 23501 8080 29524 X0888 8636 1989 51785 27538 3604 45350 25132 7633 31270 23968 7302 1990 53576 29830 3632 51785 27538 3131 26920 25142 1778 2738 81 1988 32724 20767 30320 1B127 2502 15296 10316 4980 1989 35098 23629 2798 32724 20767 2310 16375 10835 5540 1990 37902 25903 2395 35098 23629 2925 18138 11527 6611 6111 82 1988 95599 84820 89976 793 79 6293 35189 19510 15679 1989 103551 90786 6750 95599 P4820 8736 42303 21863 20440 1990 114241 98397 8147 103551 90786 11226 47523 23983 23540 1885 83 1988 22941 13532 20967 11961 2288 13644 11161 2483 1989 24110 15299 i0ß2 22941 13532 1484 12531 11449 1082 1990 30110 17027 2050 24110 15299 6322 7399 10503 -3104 1393 84 1988 3618 445 1965 1006 1711 21448 17685 3763 1989 5372 2075 725 3618 1393 1797 20626 19444 1182 1990 7475 3010 1055 5372 2075 2223 22024 20615 1409 171 771 533 05 1988 706 545 94 6330 2372 3958 168 1989 1011 646 706 545 372 7355 2743 4612 145 1990 1-313 634 1011 646 459 8309 3134 5175 2142 7988 3304 13787 5472 8315 86 1988 9955 5299 2114 9955 1989 12138 6975 1776 5299 2283 6337 5918 2419 1990 14881 8662 1855 12138 6975 2911 9531 6 311 3220 2910 10906 5670 4271 9369 4985 4384 87 1988 14810 8213 1989 13347 8750 2857 14810 8213 857 8033 4548 3485 1990 14488 10914 2561 13347 8750 1538 8389 4601 3788 726 3884 1428 88 1988 4953 2133 1090 3735 1831 1904 1989 5569 2829 743 4953 2133 663 4540 2022 2516 1990 5975 3307 486 5569 2829 414 4843 2324 2519 8294 65818 23698 89 1987 72937 33634 5477 42894 29067 13827 1988 84335 44664 10914 72937 33634 11282 44569 24532 20037 1989 92758 52589 10847 84335 44664 11345 43736 26967 16769 32033 10005 90 1987 88270 72369 25833 19706 55756 33019 2737 1988 1i X1964 35930 756 88270 32033~ 17333 69043 35839 33204 91 1987 323468 269619 21619 305132 259592 29928 92676 43965 48711 1988 346427 285859 21355 323468 269619 28074 104545 47676 56869 92 1987 69713 42293 7725 66688 39026 7483 35703 22223 13480 1988 81005 49164 9056 69713 42293 13477 37714 23099 14615 93 19B7 60432 25318 4190 54282 21218 6240 56710 18881 37829 1988 7 3336 29788 4608 60432 25318 12042 66880 21618 45262 94.1987 30668 20894 3757 26186 17436 4781 37778 25939 11839 30668 1988 34936 24379 "_B12 20894 4595 44658 29768 14890 95 1987 "26052 19806 3445 23663 16868 2696 21427 8210 13217 1988 28729 22408 3106 26052 19806 3181 26944 8832 18112 96 1987 41896 32512 4274 38681 28710 3695 20528 8806 11722 1988 45674 35175 3577 41896 32512 4692 21393 9496 11897 97 1987 50996 34927 5074 46085 30941 5999 22313 15486 6827 1988 52947 39804 5252 50996 34927 2 326 24601 1713B 7463 98 1987 28080 15523 2178 25958 13458 2235.22706 15745 6961 1988 30358 17328 1991 28080 15523 2464 24898 17557 7341 99 1987 18323 15099 2303 15151 12803 3179 12375 3034 9341 1988 23580 17943 2943 18323 15099 5356 12809 3305 9504 100 1988 549017 343225 59260 512051 293619 46620 355064 199027 156037 1989 591384 390789 51599 549017 343225 46402 373891 225169 148702 101 1988 457853 48887 50602 390209 6340 75699 146337 144634 1703 19,89 512564 79779 45753 457853 48887 69572 97451 153658 -56207 102 1988 81284 10917 9620 61988 3358 21357 51635 42170 9465 1989 102840 20409 11988 81284 10917 24052 62547 46830 15717 103 1988 '97354 39774 5196 87581 34766 9961 18724 7956 10768 1989 97911 44901 6382 97354 39774 1812 17939 9556 8383 104 1988 39311 34309 2568 38671 34017 2916 17001 12205 4796 1989 45743 35915 1973 39311 34309 6799 11241 11591 -350 105 1988 17923 13106 1301 16591 11928 1455 9272 6235 3037 1989 19549 13953 1070 17923 13106 1849 7899 7082 817 106 1988 11695 7349 1567 10756 5824 981 6141 2633 3508 1989 14234 9007 1789 11695 7349 2670 7904 3347 4557 107 1968 72 25 17 57 12 19 804 631 173 1989 174 56 34 72 25 105 883 795 Be 108 1989 56035 22349 5514 45568 16034 11666 90458 61767 28691 1990 71542 29027 7413 56035 22349 16242 94889 69793 25096 109 1989 181254 71944 19375 154162 56993 31516 65780 54995 10785 1990 216335 85013 19464 181254 71944 41476 68026 63521 4505 110 1989 92345 44215 21889 81081 19104 8042 64261 36562 27699 1994 103772 55116 13313 92345 44215 13839 61968 38243 23725 111 1989 91975 45406 11498 77846 39119 19340 46378 25966.20412 1990 122406 54596 14831 91975 45406 36072 57313 31175 26138 112 1989 4030 4011 275.2937 2897 254 22878 3902 18976 1990 4250 4243 1762 4030 4011 1750 26291 4838 21453 113 1989 187103 90467 14675 175162 83410 19559 47921 33010.. 14911 1990 204218 105345 17317 187103 90467 19554 38195 34013 4182 114 1989 38274 12628 3862 34091 9296 4713 23836 17232 6604 . 1990 47388 17359 4359 38274 12628 6742 28104 '22399 5705 115 27031 20397 3556 23 698 17636 4128 18370 10023 8347 1990 28265 20793 1747 27031 20397 2585 12912 9252 3660 116 1969 59919 34574 6461 46058 28868 14616 26943 18551 8392 1990 75016 40150 5948 59919 34574 15471 28961 20851 8110 117 1989 617 417 116 525 02 93 10038 4032 6006 1990 720 523 110 617 . 1,3417 107 13115 6261 6854 24933 15908 1823,18793 14509 118°1989 , 6564 16789 10775 6014 1990 28643 18023 2454 24933 15908 4049 15735 11686 4049 119 1989 22215 13928 1246 17641 13050 4942 14127 3937 10190 1990 27225 15530 1641 2 215 13928 5049 14650 4144 10506 120 1989 25795 15316 3626 19958 12983 7130 13019 7320 5699 1990 27856 17930 3773 25795 15316 3220 15779 8858 6921 121 1989 42472 19642 9051 41607 14196 4470 18327 7200 11127 1990 43152 24588 8654 42472 19642 4388 19310 8146 11164 122 1989 36526 15068 6007 32842 9357 3980 10904 3 568 7336 37234 19551 4643 36526 15068 868 12816 4034 8782 1990 . 123 1989 11381 8192 912 10231 7420 1290 7417 4769 2648 1990 12394 8969 797 11381 8192 1033 7001 5357 1644 124 1989 10728 5837 1748 10273 4256 622 8999 6950 2049 1990. 17556 8414 1886 10728 5837 6137 9798 7398 2400 s N ,. firm EQ. RES. L. T. F. D. loginv prorat mu lnmu no. year 200.000 47461 28656 9.76082? 1.43 7697 0.424779 0.3 54017 1 1987 22330 125440 2629B 10.00514 1.339409 0.177966 0.16"_ 789 1988 22330 109850 54582 10.28742 2.363711 0.412937 0.34567 1989 22330 1196BB 104772 11.43989 2.334783 0.737737 0.552584 1990 16000 11976 5800 9.00749 0.676059 0.207321 0.188403 2 1987 16000 8565 5781 9.751501 0.698717 0.235335 0.211342 1988 31500 1480 6203 9.516648 0.370298 0.188084 0.172342 1989 315c: 0 1335 5406 9.251002 0.249593 0.164641 -0.152413 1990 30800 20667 24524 8.733916 0.80316 0.4765 0.389674 3 1987 30BOO 31586 8323 8.952088 0.201939 0.133411 0.125232 19BB 30BOO 33094 34728 8.97866 -0.13432 0.543525 0.434069 1989 30800 30720 10784 8.503905 0.371491 0.175293 0.161517 1990 12000 23322 34000 7.321189 3.150109 0.962573 0.674256 4 1987 15000 23888 104000 7.779049 3.433688 2.674347 1.301375 19BB 15000 28685 119350 7.705713 2.90179 2.732059 1.31696 1989 15000. 31780 140B34 8.593784 3.045178 3.01056 1.388931 1990 30000 22401 25282 9.863134 1.006935 0.482472 0.393711 5 1987 1.455491 " 30000 23790 17789 10.30169 0 . 330712 0.285714 1988 30000 24680 334BB 10.03824 0.933767 0.612436 0.477746 1989 30000 24516 38603 10.07348 0.489926 0.708104 0.535384 1990 3200 14789 3189 9.118444 0.473018 0.177275 0.163202 6 1987 3200 18969 9053 9.093694 0.244701 0.408363 0.342428 19BB 3200 19896 8356 8.922525 0.445914 0.361794 0.308603 1989 3200 24155 12482 8.923724 0.574002 0.456297 0.375897 1990 27920 B5403 4059 9.657779 0.269999 0.035818 0.035191 7 1987 27920 68657 4655 9.05392 0.381405 0.0482 0.047074 1988 27920 52283 13320 9.499272 0.520335 0.166079 0.153646 1989 27920 39652 12362 10.21461 0.644531 0.182946 0.168008 1990 5000 5749 5557 8.378391 1.032055 0.516978 0.41672 8 1987 5000 6152 5366 7.878155 0.97705 0.481169 0.392832,, 1988 5000 6970 7888 7.773174. 0.987255 0.658981 0.506203 1989 5000 7710 7068 7.486613 0.687417 0.556098 0.442181 1990 30000 23910 5896 7.862112 0.843643 0.109367 0.10379 9 1987 45000 17471 4874 8.35538 0.621316 0.07802 0.075126 1988 45000 25476 23841 6.454892 0.221406 0.338285 0.291389 1989 45000 33259 21330 8.134761 -0.45239 0.272557 0.241028 1990 ' 50000 21630 14665 9. 516266 0.341965 0.204733 0.186258 10 1987 50000 20922 10864 9.33619 0.502836 0.153182 0.142525 1988 50000 27752 14575 9.588434 0.78672 0.187455 0.171812 j, 999 50000 41130 12929 9.795624 0.359503 0.141874 0.132671 1990 5700 46880 714 8.520986 4.160997 0.013579 0.013488 11 1987 5700 50783 555 8.415825 2.322813 0.009826 0.009778 1968 5700 55491 386 8.022897 1.822624 0.006308 0.006288 1989 5700 55851 1053 9.005282 1.811937 0.017108 0.016963 '1990 28500 677 40228 9.325542 0.408114 1.369371 0.862624 12 1987 60000 1077 12564 9.51856 0.326344 0.205708 0.187067 1988 32091 9.706255 0.254921. 0.51723 60000 2044 . 0.416886 1989 60000 2508 6930 9.547312 0.28043 0.110866 0.10514 1990 0-273668 22500 2665 14593 9.327768 0.579893 0.457357 `13 1967 ý' "^ f. _ A ". "t. eýH iýý, :' i. ý " '.,! ýýp ""w r ''ý. m, ,. , r , ý. 6, iýt,. ^'ý. .v , ,... .. ,'l .. .r ,_ +ý^ ,. .... _ _". .s 4500 -257 10424 8.535033 0.348388 2.456752 1.24033 1988 451.00 659 6254 7.990577 0.493931 1.21225 0.79401 1989 4500 3327 2084 8.388905 0.497954 0.266258 0.236066 1990 10000 21844 5653 9.366831 0.56 419 0.177522 0,163412 14 1987 10000 24160 5248 9.635347 1.301463 0.15363 0,142913 1988 10000 25386 4798 9.738141 1.454358 0.13559 0.127153 1989 IO Q0 45709 3176 9.760656 1.071174 0.057011 0.055445 1990 57199 100182 17270 8.777093 1.263763 0.109734 0.10412 15 1987 61645 101468 11538 8.963288 1.286433 0.070736 0.068346 1988 61645 105606 9949 9.142918 1.425141 0.059485 0.057783 1989 61645 110606 8367 9.012986 1.654484 0.046574 0.047432 1990 30000 738 5113 10.00622 0.249474 0.166341 0.153872 16 1987 30000 1448 10889 10.64132 0.065135 0.346254 0.297326 1988 15000 15000 12944 10.94706 -0.22424 0.431467 0.3587 1989 20000 0 1591 10.79144 -0.20729 0.07955. 0.076544 1990 22551 20703 7370 9.507478 0.62104 0.170389 0.157336 17 1987 22551 26682 9497 8.871084 0.6953 0.192899 0.176387 1988 22551 33199 15445 9.207336 0.607556 0.27704 0.244545 1989 22551 35897 9679 8.561593 0.558858 0.1656 0.153236 1990 42134 3462 982 8.984192 0.238985 0.021537 0.021308 16 1987 42134 5443 069 9.190852 0.138137 0.018265 0.0181 1988 47000 6595 766 9.19947 0.127066 0.014292 0.014191 1989 57000 9467 1668 9.272282 0.258987 0.025095 0.024785 1990 4038 22382 0 7.582738 2.584562 0 0 19 1987 4038 23329 0 7.277248 2.892766 0 0 1988 2736 17833 0 7.342132 3.079984 0 0 1989 2736 24856 0 7.184629 3.470627 to 0 1990 36400 45871 8922 8.960083 1.400234 0.108446 0.102959 20 1987 36400 46413 10683 8.697012 2.306202 0.129001 0.121334 1968 36400 52510 10513 9.518266 2.422433 0.118243' 0.111759 1989 36400 55137 21073 9.783972 1.696372 0.230213 0.207187 1990 27500 22069 7468 8.464003 0.466356 0.150659 0.140335 21 1987 27500 20650 6064 7.990238 0.665794 0.12594 0.118618 1988 27500 17358 5121 9.083075 0.803075 0.11416 0.108101 1989 27500 18235 3618 9.057772 0.547163 0.079108 0.076135 1990 8275 11284 31864 8.854522 0.35914 1.629122 0.96665 22 1987 13807 38116 8.952735 0.36587 1.519656 0.924122 1988 11275 . 14000 19528 32334 9.026297 0.436034 0.964388 0.675181 1989 14000 21087 31598 9.517825 0.37227 0.900561 0.642149 1990 10000 6533 0 5.869297 1.630386 0 0 23 1987 10000 4297 0 6.030685 2.566014 0 0 1,908 8990 0 6.727432 3.524238 0 0 1989 10000 . 10000 6685 0 6.466145 1.499027 0 0 1990 220 5561 0 5.513429 2.002205 0 0 24 1987 220 6243 5839 6.070738 4.223536 0.90345 0.643668 1988 220 44 0 5.942799 8.087271 0 0 1989 220 1666 34100 5.521461 5.904607 18.08059 2.948672 1990 5000 6278 2699 7.022868 1.150158 0.239315 0.214559. 25 1987 5000 6962 2362 6.753438 1.234989 0.197459 0.180201 1988 5000 7934 1986 ' 7.592B7 1.335326 0.153 549 0.142943 1989 5000 8916 1567 6.76273 0.929338 0.112604 0.106703 1990 21607 2810 9.693013 3.761991 0.09491 0.090672 26 1987 8000 2307 8000 27664 9.219994 1.867784 0.064687 0.062681 1988 8000 34994 1682 9.444463 1.854081 0.039122 0.038376 1989 8000 43672 1058 9.270212 1.593829 0.020475 0.020269 1990 28940 5859 2057 7.601402 0.290093 0.059111 0.05743 27 1987 28940 5303 18460 7.987864 0.32 00 0.539088 0.43119 1988 1 289410 6404 0 8.483636 0 0 1989 28940 7183 0 8.005033 0.39999 0 0 1990 2000 14983 17164 8.642239 0.527332 1.010658 0.698462 28 1987 2000 14491 732 8.946245 0.45707 0. i X44388 0.043431 1988 2000 15763 1582 9.04204 0.457391 0.089062 0.085316 1989 2000 17612 3064 9.562756 0.468223 0.156231 0.145165 1990 1000 1902 1694 7.659171 2.8224 0.583735 0.459786 29 1987 1400 3155 2182 8.0 0599 3.137679 0.52515 0.42 2093 1988 5000 2084 2727 8.101678 1.748484 0.384952 0.325665 1989 5000 2509 2727 8.607399 0.522239 0.363164 0.309809 1990 30 7000 56359 25600 9.9 4662 0.138132 0.404047 0.339359 1987 7000 47830 23011 10.43079 0.535618 0.419679 0.3 50431 1988 7000 54742 20253 10.90099 0.765179 0.328026 0.283694 1989 7000 66049 25423 10.5052 0.3603 97 0.348027. 0.298642 1990 15000 190 9917 8.341887 0.407853 0.652864 0.502509 31 1987 15000 1019 8703 8.542861 0.38617 0.543292 0.433918 1988 15000 1802 6898 7.8493 24 0.377873 0.410546 0.34 3977 1989 16500 2312 5773 6.781058 0.36773 4.306879 0.267642 1990 7.401231 5000 -1331 19202 0.435697 5.233579 1.829951 32 1987 5000 2378 8032 7.586804 0.403843 1.088642 0.736514 1988 8000 2725 8190 8.376321 0.432709 0.763636 0.567378 1989 8000 3792 7687, 8.721113 0.186181 0.651543 0.50171 1990 33 4000 7755 2533 7.746301 1.403845 0.215483 0.195141 1987 4000 8681 2474 7.833204 1.702945 0.195095 0.178226 1988 4000 13964 1877 7.627544 1.882825 0.104487 0.099381 1989 10000 2110 28374 10.39259 2.383582 2.343022 1.206875 1990 10000 50799 3857 8.919721 1.367841 0.063439 0.061508 34 1987 10000 53013 3267 8.395929 1.16983 0.051846 0.050547 1988 10000 59122 2360 8.198914 0.899205 4.034143 0.033573 1989 10000 69914 432 8.3 4231 0.678859 0.045406 0.005391 1990 2400 3870 0 5.863631 1.14003 a 0 35 1987 4800 2379 0 5.762051 1.743881 0 0 1988 4800 3661 0 5.976351 2.076685 0 0 1989 4800 5370 0 5.505332 2.136132 0 0 1990 650 326 0 5.459586 0.441303 0 0 36 1987 650 503 0 5.141664 0.251853 0 0 1988 650 263 0 4.317488 0.183752 0 0 1989 650 313 0 4.26268 0.231072 0 0 1990 9000 849 3014 8.040447 0.47497 0.306021 0.266985 37 1987 9000 1564 9481 9.957407 0.571915 0.89 7482 0.640528 1988 9000 4807 15613 9.131405 0.414633 1.1 4803 0.756499 1989 9000 9702 7301 8.084871 0.502799 0.390386 0.329581 1990 13003 3411 9.018332 0.464009 0.213148 0.193218 38 1987 3000 0.34 15187 2239 8.576405 362 0.122103 0.115204 1988 3150 . 16683 10098 9.356689 0.455819 50533 0.409012 1989 3300 -. 3940 20747 13235 9.455793 0.433976 0.5 3611 0.429255 1990 5780 1765 10061 9.111293 0.967486 1.333466 0.847355 39 1987' 5780 280') 7220 8.080547 0.491504 0.841492 0.610576 1988 5780 2183 11256 7.891705 0.653185 1.413538 0.881094 1989 7000 2249 11773 8.262483 0.64795 1.272894 0.821054 1990 22230 2442 5617 7.532088 0.557767 0.227667 0.205116 40 1987 22230 2253 5000 6.971669 0.674039 0.204223 0.185835 1988 22230 3465 5700 7.45068 0.737503 0.221833 0.200352 1989 22230 3833 6290 7.127694 0.763989 0.241338 0.21619 1990 4000 3873 7008 5.937536 '0.56179 0.890131 0.636646 41 1987 ý.. ý, irr, ý! .. 4ýý _ :n ý'F'ý . ,.. rýý. "" _ ý"iýS; 4000 4447, 5700 7.691657 0.721484 0.675115 0.515882 1988 4000 4546 7200 7.95367 0.821107 0.842499 0.611123 1989 4000 4939 6718 7.597898 0.671153 0.751538 0.560494 1990 1100 13295 2404 8.089482 2.555365 0.167002 0.154438 42 1987 1100 15987 2277 8.927181 3.809512 0.133259 0.125098 1988 1100 19887 2138 9.45313 1.536938 0.101873 0.097011 1989 11OQ 21212 3850 9.048174 1.068681 0.172553 0.159183 1990 6600 189 0 7.244228 0.548666 0 0 43 1987 6600 2642 4950 7.434257 0.696076 0.535598 0.42892 1988 6600 5691 0 9.18338 0.856654 0 0 1989 6600 8024 5000 8.602269 0.335253 0.341904 0.294089 1990 15190 11877 12451 8.726968 0.312477 0.460007 0.378441 44 1987 15190 11940 14571 8.50593 0.363281 0.537081 0.429885 -- 1988 15190 12129 17594 8.273081 0.529776 0.644021 0.497145 1989 15190 12465 15330 5.697093 0.448726 0.55433 0.441045 1990 10400 1775 7000 6.89467 0.804896 0.574949 0.454223 45 1987 10400 3103 7000 7.4313 0.817775 0.518403 0.417659 1988 10400 3960 6500 7.005789 0.6228 0.452646 0.373387 1989 10400 5189 5500 8.144098 0.669544 0.352813 0.302186 1990 2600 4294 8192 8.434464 1.094475 1.18828 0.783116 46 1987 2600 5640 10698 7.351156 0.676702 1.267536 0.818694 1988 2600 7149 10347 7.41998 0.826483 1.06124 0.723356 1989 2600 8747 9911 8.244071 1.486976 0.873447 0.62778 1990 3200 2001 7,008 7.444249 1.490383 0.57835 0.45638 47 1987 3200 3519 2463 7.583756 1.36065 0.366572 0.312306 1988 3200 4999 3044 7.803435 1.571576 0.371265 0.315734 1989 3200 7037 2469 8.737934 1.514404 0.241184 0.216066 1990 1000 16481 13066 7.913887 0.393162 0.74744 0.558152 48 1987 1000 16752 5868 7.822445 0.440199 0.330554 0.285596 1988 1000 17841 6143 8.156223 0.382008 0.3 26044 C, 28222 . 1989 1000 22150 0438 8.80807 0.561745 0.364492 0.310783 1990 250C) 1684 0 7.367077 2.258317 0 0 49 1967 2500 2621 0 7.650645 2.111646 0 0 1988 2500 3862 0 7.287561 1.695394 0 0 1989 2500 4735 0 7.580189 1.393036 0 0 1990 4285 5597 6370 8.360539 0.4; 9722 0.644606 0.497501 50 1987 4285 5643 5866 9.049819 0.24119 0.590854 0.464271 1988 4285 5645 8749 9.466532 0.123133 0.881067 0.631839 1989 4285 3483 12028 6.946976 -0.06716 1.548404 0.935467 1990 12000 38 5608 7.763871 0.313238 0.465858 0.382441 51 1987 12000" 662 5155 9.016634 0.505551 0.407124 0.341548 1988 12000 2170 4122 8.560061 0.399905 0.290896 0.255337 1989 12500 3050 2969 8.232706 0.281268 0.190932 0.174737 1990 2800 696 3788 7.160069 0.375265 1.083524 0.734061 52 1987 2800 904 3O2 7.085901 0.469486 0.891469 0.637354 1988 . 2 800 963 5353 7.130899 0.382814 1.422535 0.864815 1989 2800 1028 5925 6.652863 0.515864 1.547806 0.935232 1990 2000 1244 0 3.850148 128.7508 0 0 53 1987 2000 1.95 0 4.727388 45.04247 0 0 1988 2000 -317 0 5.023881 21.11866 0 0 1989 2000) 304 0 6.173786 17.64825 0 0 1990 14C)C) 3787 3198 7.325808 0.557917 0.616541 0.4802 89 54 1987 1400 3728 4183 8.442901 0.335495 0.815718 0.596481 1988 2516 11553 21253 10.62147 0.290795 1.510626 0.920532 1989 5816 11653 4912 7.038784 0.0442 68 0.281184 0.2A. 7785 1990 8400 13759 10380 9.931297 1.527004 0.468433 0.384196 55 1987 15240 9581 8.898912 0.339875 0.405288 0.340242 1988 16106 11493 7.177782 0.358272 0.468987 0.384573 1989 15893 9993 6.562444 0.168004 0.377194 0.320048 1990 762 1622 8.033983 0.167451 0.049509 0.048322 56 1987 52*2 1291 8. 2 36421 0.223597 0.039696 0.038929 1988 987 3056 8.646114 0.176759 0.09 643 0.088599 1989 1379 2727 8.3 69621 0.098333 0.081698 0.0785 32 1990 2821 1409 6.2 4558 1.557124 0.363051 0.33097 5 57 1987 3-540 1279 6.124683 1.769148 0.278043 0.24533 1988 3 718 1250 6.251904 1.26009 0.261616 0.232393 1989 4853 2250 6.695799 0.437423 0.380518 0.32 458 1990 1501 0 5.361 92 2.636686 0 0 58 1987 1196 0 6.721426 3.475027 0 0 1988 1678 0 6.788972 2.481856 0 0 1989 1835 0 5.66296 2.874933 . 0 .0 1990 -3212 0 7.004882 0.015056 0 0 59 1987 -3579 0 7.669962 -0.29483 0 0 1988 -7087 0 7.550135 1.418442 0 0 1989 -5695 0 8.63693 1.042691 0 0 1990 3522 974 7.641564 0.840856 0.068389 0.066152 60 1987 5383 1987 7.95 263 0.577641 0.123393 0.116354 1988 6123 7714 8.187855 0.393908 0.457994 0.377062 1989 6620 742; 1 8.055475 0.260712 0.42797 0.356254 1990 4785 0 8.500454 0.784293 0 0 61 1987 7022 0 9.276783 0.333721 0 0 1988 9044 0 10.21618 0.349762 0 0 1989 11106 0 9.158205 0.194095 0 0 1990 24067 28918 10.07778 0.485636 0.200726 0.182926 62 1987 38661 33318 10.18867 0.617421 0.209995 0.190616, 1968 57716 32581 10.58777 0.482511 0.183332 0.168334 1989 98861 3379 9.490998 1.573048 0.030757 0.030294 63 1987 124077 20947 9.4796194 1-. 60715B- 0.155075 0.144165 19BB 152722 23137 9.775597 1.314802 0.141319 0.132184 1989 86560 51298 9.973946 0.056773 0.531255 0.426088 64 1987 -5471 85597 10.23635 -0.01392 0.90551 0.64475 1988 9.887612 -29565 108092 -0.10142 1.534635 0.93005 1989 54288 10901 5.7 6848 0.824092 0.17501 0.161276 65 1987 15067 9386 8.116417 1.225326 0.148121 0.138127 1988 15041 7762 8.446771 1.21234 0.122543 0.115597 1989 14445 0 7.069874 11.6310 0 0 66 1987 16637 0 6.463029 12.55741 0 0 1988 18510 0 6.874198 16.55937 0 0 1989 2782 67551 10.81968 0.187711 14.1261 2.716422 67 1987 1017 105586 10.4765 0.194979 34.99702 3.583436 1988 1218 71948 10.13805 0.208213 22.35799 3.150939 1989 3257 11679 8.135054 0.451135 1.26164 0.81609 68 1987 4410 14341 8.644002 0.272527 1.377618 0.866099 1988 2349 16624 8.20166 0.22501 1.991137 1.095653 1989 33906 45741 8.557183 0.060724 0.935284 0.660254 '69 1987 92691 29949 10.34856 0.479534 0.278101 0.245376 1988 88198 13454 8.900004 0.252221 0.130371 0.122546 1989" 12708 193 7.003974 2.735142 0.014032 0.013935 76 1987 15105 153 7.03966 2.636359 0.009473 0.009429 1988 16515 748 7.658228 2.532174 0.042594 0.041712 1989 7389 1093 7.823646 0.992517: 0.086822 0.083258 71 1987 8162 1367 7.762596 0.865962' 0.102305 0.097403,. - 1989 520ä 03.9 1367 7.734559 1.148258 0.100968 0.096189 1989 2700 3409 2716 7.07327 1.337882 0.44459 0.367826 72 1987 2700 3522 3820 7.42 9521 0.529296 0.61395 0.478685 1988 2700 3693 3529 7.665572 0.628353 0.55201 0.439551 1989 3000 1226 2479 7.746029 0.310547 0.586607 0.461590 73 1967 3000 1233 1577 7.639642 0.259543 0.372549 0.31667 1988 3000 -132 2754 7.993958 0.256241 0.960251 0.673073 1989 2000 4196 0 7.431892 0.817873 0 0 74 1987 2000 5657 0 6.806B29 1.251203 CU 0 1988 2000 5B23 0 7.640123 1.777313 0 0 1989 400 0 8895 1040 7.926964 -0.. 35891 0.02127 0.021047 75 1987 40000 100215 1526 8.19257 -0.54212 0.030389 0.029937 1988 40000 6503 1375 8.480944 -0.36606 0.02 9568 0.02 9139 1989 1015 256 0 3.871201 28.72759 0 0 76 1988 1015 1387 0 3.806662 19.54835 0 "0 1989 1015 2937 0 3.951244 22.14195 0 0 1990 4000 27389 1584 8.606302 0.598708 0.050464 0.049232 77 1988 4000 30065 -757 8.955964 0.555903 0.022222 0.021979 1989 4000 33951 1085 8.989195 0.528486 0.028569 0.028168 1990 17715 3944 9589 7.721792 0.801016 0.442726 0.366534 78 1988 17715 18542 8457 9.025576 0.64435 0.233252 ((. 209654 1989 17715. 26787 7132 7.997327 0.367288 0.160262 0.148646 1990 10000 1445 13875 8.122668 0.246071 1.21232 0.194042 79 1988 15000 9615 5296 8.594895 0.297411 0.215153 0.19487 1989 25000 1888 10762 9.167015 0.342903 0'. 400253 0.336653 1990 10200 10732 5893 8.997147 0.534643 9.281531 0.248055 80 1988 20200 1346 3482 8.940236 0.34021 0.161608 0.149805 1989 20200 3284 2420 8.049108 0.069512 0.103049 0.098078 1990 8000 12676 4228 7.824846 0.368738 0.204486 0.186055 81 1988 8000 12770 2734 7.745003 0.436446 0.131632 0.123661 1989 8000 12811 5817 7.98105 0.546425 0.279516 0.246482 1990 2300 16637 5486 8.747193 1'. 408229 0.289697 0.22+4408 82 1988 2300 42304 7870 9.075208 1.786264 0.176442 0.162494 1989 2300 51369 6401 9.325988 1.748133 0.119268 0.112675 1990 210 4010 0 7.735433 0.262411 0 0 83 1988 4000 20191 0 7.302496 0.108325 0 0 1989 4000 20965 0 8.751791 -0.33395 0 0 1990 7500 25842 0 7.444833 3.73468 0 0 84 1968 7500 28830 0 7.493874 0.500415 0 0 1989 7500 28690 0 7.706613 0.405117 0 0 1990 1000 2834 13000 4.543295 15.82839 3.390715 1.479492 85 1988 1000 3f) 99 13000 5.918894 26.98402 3.171505 1.428277 1989 1000 3145 20000 6.12905 13.44021 4.82509 1.762175 1990 3500 974 1935 7.656337 1.689597 0.432499 0.35942 86 1988 350') 1060 2229 7.733246 0.489402 0.488816 0.397981 1989 3500 1431 1705 7.976252 0.591211 0.345772 0.296968 1990 3300 415 2011 8.359603 0.796909 0.541319 0.432639 87 1988 3300 762 2934 6.753438 0.497622 0.722304 0.543663 1989 3300 1743 6141 7.338238 0.781132 1.217728 0.796483 1990 75tß 1384 2469 6.993933 0.737864 1.156992 0.76871 88 1988 1206 790 3580 6.496775 0.841104 1.793597' 1.0273 26 1989 ' 1350 913 4326 6.025866 0.871498 1.911622 1.06871 1990 7000 4923 25013 8.608313 0.309307 2.097878 1.130717 89 1987 7000 5305 17857 9.330964 0.485227 1.451199 0.896577 1988 7000 6117 14530 9.336532 0.398178 1.107723 0.745608 1989 10,00 17612 5181 9.888678 0.460356 0.185619 0.170265 '90 1987 . ýý S 318944 10::. i c) 21251 10063 9.760367 0.561961 00 0.276831 1988 24396 103608 34769 10.30655 1.007821 0.271624 0.240295 91 1987 2457.3 115758 29387 10.2426 1.005161 0.209412 0.190134 1988 12000 14422 28514 8.920389 0.459151 1.079176 0.77,1972 92 1987 1 000 11220 28256 9.50874 0.5073,05 1.216882 0.796102 1988 1286 25294 667 8.738735 1.078001 0.025094 0.024784 93 1987 1286 25397 242 9.396156 1.226849 0.009069 0.009029 1988 3024 8530 8833 8.472405 1.274842 0.764497 0.567866 94 1987 12945 9746 8.432724 1.449974 0.610307 0.476425 1988 .=7024 5289 1272 7.899524 1.780306 0.095718 0.09141 8000 . 95 1987 7400 83-12 269 8.064951 2.759956 0.017121 0.016976 1988 1800 22883 335 8.214736 1.108565 0-013572 0.013481 96 1987 "1800 23235 0 8.453614 1.206667 0 0 1988 7500 7679 7736 8.699348 0.424755 0.509651 0.411879 97 1987 7500 9642 6859 7.751905 0.442041 0.400128 0.336564 1988 23000 3724 4003 7.711997 0.5247 0.14979 0.13958 98 1987 23000 3821 4124 7.809541 0.556426 0.15376 0.143026 1968 4000 6487 2612 8.064322 3.748391 0.24907 0.222399 99 1987 4000 6094 6216 8.585973 2.805751 0.615811 0.479837 1988 242825 64995 67845 10.74978 0.679906 0.220405 0.199183 100 1988 242825 67889 63210 10.7451. 0.680662 0.203435 0.18518 1989 95000' 234 225358 11.23452 0.004222 2.366361 1.21 3832 101 1988 76993+ 223217 11.15012 2.037488 11.111031 1989 . 32562 -0.12946 75015 0 4172 9.969135 0.153652 0.055616 0.054124 102 1988 55225 0 10.08797 0.210399 0 1989 9020 .0 7000 14487 25553 9.206433 0.194051 1.189231 0.78355 103 1988 7000 16794 24743 7.502186 0.137142 1.039884 0.712893 1989 2000 23361 106 7.977968 0.984823 0.00418 0.004171 104 1988 2000 24494 3300 8.824531 -0.06591 0.124557 0.117389 1989 1500 4762 656 7.282761 0.619894 0.104759 0.099627 105 1988 1500 5116 516 7.5224 0.15976.8 0.077993 0.075101 1989 1750 1812 4598 6.888572 0.676978 1.290848 0.828922 106 1988 1750 2784 3852 7.889834 0.987717 0.849581 0.614959 1989 200 2 0 2.944439 3.659076 0 0 107 1988 200 6 0 4.65396 1.763713 0 0 1989 32000 23752 342 9.364434 0.981566 0.006134 0.006116 108 1989 . 32000 31164 212 9.695356 0.706226 0.003Z56 0.003351 1990 .80000 -10845 18582 10.35825 0.104553 0.268701 0.237993 109 1989 80000 -18158 19997 10.63287 0.039068 0.323356 0.280171 1990 58516 -4899 10659 8.992433 0.420995 0.197693 0.180397 110 1989 58816 -5956 376 9.535246 0.467282 0.007113 0.007088 1990 40000 3303 33507 9.869931 0.496495 0.77378 0.573113 111 1989 55000 3719 43087 10.49327 0.532064 0.733783 0.550306 1990 8500 12347 0 5.537334 446.8769 0 0 112 1989 8500 89? 5 0 7.467371 1070.343 0 0 1990 5,235 802 49326 9.881191 0.153086 0.912819 0.646578 113 1989 53235 -1665 59185 9.880935 0.041024 1.147663 0.76438 1990 27000 3116 3812 8.45808 0.250892 0.126577 0.119184 114 1989 27000 3132 4168 9.075894 0.210875 0.138325 0.129558 , 1990 3000 11827 3292 8.325548 1.297053 0.222027 0.200511 115 1989 40017 12766 2897 7.857481 0.522941 0.17279 0.159386 1990 13500 3620 5467 9.589872 0.459868 0 319334' 0.277127 116 1989 . . 13500 6364 4980 9.646723 0.303331 0.25070 5 0.223707 1990 2000 1669 0 4.532599 25.37019 .U 0 117 1989 2000 2311 0 4.672829 32.48649 0 0 1990 5583 1687 E3.789355 J. 322: 387- 0-162165 4820 - 0.150284. 118 1989 4820 4659 4387 8.306225 0.425294 0.462013 0.380361 1990 750() 8960 2745 8.505525 2.090788 0.166768 0.154237' 119 1989 7500 10754 2214 8.5 6945 1.20179 0.121288 0.114478 1990 50000 9810 9587 8.872067 0.769658 0.647333 0.499158 120 1989 5000 10445 8756 8.077137 0.6 6091 0.566915 0.449109 1990 990C) X176 73095 8.405144 0.382381 5.590012 1.885555 121 1989 9900 5896 63784 8.386629 0.463556 4.037984 1.617006 1990 6175 13367 3318 8.289037 0.294247 0.169788 0.156823 122 1989 6175 13556 6000 6.766192 0.307965 0.30409 0.265505 1990 10010 5726 2402 7.162397 0.887361 0.357122 0.3053 66 123 1989 1000 6646 2948 6.940 22 0.468693 0.365561 0.326105 1990 1325 613 2111 6.43294 0.320778 1.089267 0.736813 124 1989 1325 624 5566 8.722091 0.46516 2.855823 1.349585 1990 [, ' 1 :; Electroniccsector firm t-1 no. year 1< ACC. D. DEFT. K(-1) ACC. D. INV. V. ADD. LAB. C. 4494 2.5 223714 57351 409375 1 1987 . 188161 61848 510511 342897 1988 510760 277393 6915 449425 223714 76871 605400 382538 1989 599909 344675 75969 510760 277393 97836 754469 419015 1990 673979 423489 91170 599909 344675 86426 776516 462490 2 1987 369245 323241 4976 358466 297459 34759 361234 24064~ 1988 445071 357991 56925 369245 323241 98001 394982 265317 1989 615911 391448 98397 445071 357991 235780 423720 30496 £990 629921 419625 111814 615911 391448 97647 307051 :329869 3 1987 135359 92 271 13368 123 318 82128 15266 122768 85110 1988 146411 99: 328 10316 135359 14311 122710 92271 .,. 8767 1989 157340 104973 10560 146411 99328 15844 146578 98~3-5 1990 166218 112659 11625 157340 104973 12817 165046 104724 4 1987 59379 36541 32444 43307 24539 36514 170737 131389 1988 92400 42679 10961 59379 36541 37844 164294 105699 1989 139482 57436 19161 92400 42679 '51486 185469 130977 24021 1990 167816 74368 139482 57436 ;.~542313413 149039 5 1987 21137 8619 1689 20560 7997 1644 46328 20226 1988 25849 10+15 2060 21137 8619 5076 56037 22497 1989 26708 11661 2405 25849 10315 1918 63336 25582. 1990 31445 14657 3812 26708 11661 5553 61488 29880 6 1987 193897 154949 22649 172224 136266 25639 184570 145836 1988 223405 184809 30396 193897 154949 30044 266650 199065 1989 252123 208302 28955 223405 184809 34180 337067 215274 1990 279549 230976 30102 252123 z08302 34854 410189 230934 7 1987 90753 65720 11019 86761 55745 5036 61567 6863 1988 101097 71911 9486 90753 65720 13639 125812 80495 1989 167042 80432 9558 101097 71911 66982 104748 93843 1990 449940 85143 19057 167042 80432 297244 63109 102947 8 1987 95190 43966 17925 84847 35529 19831 119478 106938 1988 108788 54073 20861 95190 43966 24352 136377 1: 0751 1989 125377 76111 31631 108788 54077, 26182 158776 148861 1990 146958 88948 14975 125377 76111 23719 168137 168209 9 1987 161614 146200 16094 173636 154105 11977 129912 70965 1988 147259 127343 10529 161614 146200 15031 104603 78797 1989 135976 115514 1010 147259 127343 1556 89058 86492 1990 122763 90880 5957 135976 115514 17378 64214 94706 10 1987- 4090 2949 1113 3355 2104 1003 5439 44610 1988 4660 :3702 1091 4090 2949 908 65027 . 55977 1989 5265 4358 1255 4660 3702 1204 77289 68779 1990 5673 4539 1425 5265 4358 1652 91484 83773 11 1987 51018 31623 6670 44291 26069 7843 89064 74754 1988 59381 37957 6853 51018 31623 8882 101131 80421 1989 64980 44187 7028 59381 37957 6397 121228 93540 1990 69134 48287 6431 64980 44187 6485 141362 103907 12 1987 143830 77160 23426 116698 62291 35689 98848 45300 1988 178364 98136 39736 143830 77160 53294 112506 52443 1989 227126 122025 41794 178364 98136 66667 147640 63816 1990 257923 146238 44122 227126 122025 50706 159351 75783 122365 64914 66034 121082 13 1987 60170 6573 -3118 56308 1988 125627 71248 24515 122365 64914 21443'81240 53 543 1989 141978 77953 16490 125627 71248 26136 74963 55071 1990 146097 83268 15278 14.978 77953 14082 76013 55385 14 1987 43515 25370 5722 32223 20218 11862 102708 78252 1988 66849 32713 9619 43515,25370 . 25610 129108 95710 1989 76510 42866 13 536 66849' 32713 13044 164261 126012 1990 82197 57480 19008 76510 42866 10081 194076 146920 15 1987 74046 46016 17850 73158 43082 15804 100375 66702 1988 80855 50449 1809, 74046 46016 20468 106800 75714 1989 90292 55219 14536 80855 50449 19203 111923 84889 1990 100459 63842 12193 90292 55219 13737 113001 93069 16 1987 8118 3082 2879 7056 2505 3364 14566 10145 1988 9390 4186 1236 8118 3082 1404 26408 1 184 5322 1989 11706 1256 9390 4186 2436 29643 16950 1990 14606 5964 1686 11706 5322 3944 46474 24427, 17 1987 92654 33767 19082 59937 25508 43540 69131 50404 1988 9630 42823 17471 92654 33767 12083 76140 53228 1989 103141 52803 16847 96322 4 823 13686' - 88303 61191 1990 111154 62202 16 334 103141 52803 14948 95386 71564 18 1987 8351 4839 1458 6883 3567 1654 26836 10208 1988 11945 6090 1571 8351 4839 3914 30565 11679 1989 13679 7661 1952 11945 6090 " 2115 41800 14354 1990 17088 9072 2256 13679 7661 4254 45165 16301 19 1987 63726 45430 8140 53893 37633 10176 57242 44064 1988 72241 M833 8477 63726 45430 9589 62985 47172 1989 114592 58890 5883 72241 52833 42177 42331 3 782 1990 122834 68573 12719 114592 58890 11278 71910 56703 20 1987 15197 7626 2955 10268 4937 5195 15690 5303 1988 23979 10979 4045 15197 7626 9474 20237 7304 1989 26721 15106 4874 23979 10979 3489 20277 10944 1990 35481 19060 5097 26721 15106 9903 27924 11046 21 1987 31162 16285 2743 28395 '14401 3626 37698 35078 1988 33719 18449 2966 31162 16285 3359 42219 37189 1989 38443 20786 4252 33719 18449 6639 42869 42004 1990 42924 23907 5972 38443 20786 7332 46754 45871 . 22 1987 2960 2168 479 2749 1720 242 30064 16210 1968 3328 2495 453 2960 2168 494 34021 18311 1989 3757 2823 450 3328 2495 551 34801 22353 1990 4037 2901 430 3757 2823 632 32703 22740 23 1987 45178 21932 5982 40425 17339 6142 40957 32086 1988 48205 29509 7865 45178 21932 3315 48139 36572 1989 52468 33589 4866 48205 29509 5049 50920 42449 1990 54259 37579 4487 52468 33589 2288 55927 47256 24 1987 18441 9895 1457 16602 8593 1994 28300 15161 1988 20616 10807 1566 18441 9895 2829 32367 17035 1989 22316 12005 1870 20616 10807 2372 37783 19726 1990 24254 13544 2216 22316 12005 2615 35619 22332 25 1987 54007 26273 8049 34341 18904 20346 30102 19417 1988 "-75114 33834 9163 54007 26273 22709 40617 26893 1989 78906 4 101 13949 75114 33834 9474 57721 39677 1990 102716 58741 20635 76906 42101 28005 76909 50517 26 1987 16065 4267 2415 9808 1280 5685 14501 10968 1988 23294 9828 5634 16065 4267 7302 28862 24402 1989 29862 16505 6794 23294 9828 6685 37298 28311 1990 39601 24015 7583 29862 16505 9812 42541 ; 1232 27 1987 73312 48057 9335 64048 42791 13333 2712 0 1968 89938 56077 9964 73312 48057 18570 6795 0 1989 102861 64403 10924 89938 56077 15521 8478 0 1990 111695 71101 11519 102861 64403 13655 9100 0 28 1987 8191 6650 1955 692.1 5062 1637 26034 11174 1988 9417 7939 1857 8191, 6650 1794 32213 13244 14722 9041 1538 9417 1989 , 7939 5741 23174 15353 16767 ' 1880 14722 1990 28761 9041 4205 17542 16647 29 1987 4162 1469 236 4017 1387 299 6245 2602 1988 4433 1744 350 4162 1469 346 9787 3108 1989 4974 2059 436 4433 1744 662 10826 3621 1990 5096 2158 338 4974 2059 361 8085 4676 1987 36944 20577 3362 29228 18122 8623 21994 20178 30 22994 1988 39388 3216 36944 20577 3243 28070 20983 1989 46291 27680 5055 39388 22994 7272 40130 24151 Ago 48906 31852 5552 46291 27680 3995 34671 25997 31 1987 23994 15067 2632 20432 12542 3669 13224 9844 1988 26712 17839 2778 23994 15067 2724 12746 9767 1989 29749 20228 2959 26712 17839 3607 15009 11971 1990 32859 23141 3382 29749 20228 3579 16476 13549 32 1987 6234 2651 1634 5770 1935 1382 11778 7790 1988 7136 3563 1816 6234 2651 1806 17847 10404 1989 8480 4546 1379 7136 3563 * 1740 17996 11179 1990 10333 5584 2000 8480 4546 2815 21556 13410 33 1987 5 219 4010 691 4533 3353 720 13014 6738 1988 6197 4647 684 5219 4010 1025 13710 8048 1989 6444 5037 832 6197 4647 689 14663 10059 1990 6434 5443 793 6444 5037 377 16627 10131 28350 34 1987 30057 19911 6744 29847 17424 4467 30662 1988 30872 19549 7994 30057 19911 9171 29995 28054 1989 31230 20879 11472 30872 19549 10500 28894 30240 1990 26168 16329 7979 31230 20879 7467 26857 25198 35 1907 11161 8648 1793 9669 7133 1770 18882 11898 1988 13330 10207 1941 11161 8648 2551 24753 15668 1989 17757 12465 2443 13330 10207 4612 30839 19086 1990 28444 16200 3815 17757 12465 10767 48570 31700 36 1987 8928 5383 1546 6613 4190 2668 22941 10482 1988 12283 7010 1949 8928 5383 3677 23959 13302 1969 16097 9232 1987 12263 7010 3579 19258 14095 1990 18522 10773 1626 16097 9232 2510 21138 14809 37 1987 4413 2402 710 3930 1727 518 14681 8872 1988 5151 3091 733 4413 2402 782 16682 10001 1989 5309 3263 626 5151 3091 612 19359 11929 1990 6094 4009 791 5309 3263 830 22777 13725* 38 1987 8520 7237 1876 7662 5789 1286 16836 12130 1988 9307 8286 1329 8520 7237 1067 16634 13204 1989 9228 8309 1090 9307 8286 98B 19221 15124 1990 9516 8758 1039 9228 8309 878 19847 16009 39 1987 2458244 2262451 538116 2135121 1661451 266239 22587.40 932657 1988 2648774 1807606 466316 2458244 2262451 1111691 2623903 1024749 1989 2860550 1982834 445856 2648774 1807606 482404 2760068 1156572 40 1987 70538 44017 6633 58834 39195 13515 188514 116007 1988 83073 50023 8060 70538 44017 14589 212346 129100 1989 92877 59820 11215 83073 50023 11222 270354 145123 41 1987 341171 191912 27361 323391 167833 21062 652073 272544 1988 378806 216324 28924 341171 191912 42147 392330 301713 1989 424105 246002 33214 378806 216324 48835 476602 339577 1561 765 424 1010 412 642 6550 1436 42 1987 3351 1988 1202 413 1581 765 1746 11700 3241 1989 5861 1763 588 ' 3351 1202 2537 11375 4576 43 1988 227229 50064 34761 172378 21381 60929 227515 178130 1969 249837 78622 42064 227229 50064 36114 216967 185164 1990 273248 120138 50898 249837' 78622 32793 194863 171301 122193 23647 '127979 '44 1988 165701 102047 41223 273917 86766 1989 203978 148067 31175 165701 122193 43578 304472 104693 1990 240757 178884 35623 203978 148067 41585 324652 115991 45 1988 1554 793 445 1135 588 659 22475 3264 1989 1744 1173 546 1554 793 356 26102 4004 1990 1841 1285 487 1744 1173 472 32670 4936 46 1988 21932 13522 3588 25474 14772 1296 27649 21002 1989 25207 16730 5187 21932 13522 5254 29957 21673 ; 1990 29885 20710 5008 25207 16730 5706 30193 23611 47 1988 34271 27338 2524 31523 25250 3184 41729 26698 1989 48433 27285 2172 34271 27338 16387 46472 32227 1990 54019 29459 2809 48433 27285 6221- 47892 36175 48 1988 7433 5146 1458 6771 4352 1326 10930 5736 1989 8825 5558 1366 7433 5146 2346 9458 5786 1990 9329 5701 1273 8825 5558 1634 8916 6400 49 1988 10940 4176 1242 8934 3434 2506 15161 13145 1989 12310 5191 1456 10940 4176 1811 17598 14712 1990 14028 6335 1569 12310 5191 2143 20024 16142 50 1988 13996 2088 1121 7026 981 6984 9804 10787 17347 4658 2045 13996 2088 2826 16471 15119 1989 . 1990 19331 6556 -2686 17347 4658 2772 17508 15765 51 19BB 2853 1678 1526 2796 1090 995 6595 3526 1989 2966 1862 1127 2853 1678 1056 6031 3746 1990 3015 2035 1116 2966 1862 992 5580 4113 52 1988 3426 2777 153 3168 2669 283 6042 3872 1989 3840 2952 296 3426 2777 535 6581 4452 1990 4631 3447 504 3840 2952 800 7540, 4839 53 1987 175806 104242 15014 160672 94570 20476 182136 180631 1988 198885 115513 16602 175806 104242 28410 246812 191438 54 1987 54511 35401 9023 47220 28569 9482 83208 59880 1988 65622 40933 8778 54511 35401 14357 99315 70610 55 1987 45197 32749 5007 43343 29988 4100 53674 39631 1988 46532 35692 4755 45197 32749 3147 61048 45455 56 1987 32510 21603 4501 34633 21854 2629 29170 22413 1988 30963 21398 3519 32510 21603 2177 13587 18057 57 1987 26627 16618 4823 22167 14914 5579 21085 8157 1988 37238 23424 5541 26627 18618 11346 28969 10098 58 1987 4586 3133 986 3821 2520 1138 24513 7370 1988 8693 4191 10101 4586 3133 13150 29455 9532 59 1987 13161 9984 1305 12081 8763 1164 16762 10810 1988 15190 11318 1514 13161 9984 2209 19797 12340 60 1987 14354 8319 1593 12648 6758 1738 14389 12236 1988 16377 9843 1720 14354 8319 "2219 17609 13867 61 1989 326894 229720 47908 275513 199795 69364 486674 252432 1990 413404 267947 55071 326894 229720 103354 457342 281556 62 1989 404961 109033 62791 354045 53188 57862 195060 207266 1990 509202 179724 82666 404961 109033 116216 267218 221965 63 1989 29395 11787 5553 8736 2486 16911 60720 13295 1990 47215 20116 10598 29395 11787 20089 123572 '25810 64 1969 34232 17923 10419 27906 7724 6546 51653 '16343 1990 37121 24506 6854 34232 17923 3160 39832 31470 65 1989 46646 18839 9306 41398 11521 7236 79219 51056 1990 50008 23197 5726 46646 18839 4730 58328 35298 66 1989 10831 6402 5930 12462 4276 2173 53756 42687 1990 12462 8498 8966' 10831 6402 8501 50774 39341" 67 1989 14935 7965 3444, 10353 5197 5258 33929 21651 4102 1990 19284 11359 14935 7965 5057 -46142 25346 68 1989 14799 6458 2552 11160 5293 5026 23763 14129 1990 15837 7945 3237 14799 6458 2788 23061 16065 69 1989 10583 5008 3287 7123 2432 4171 18097 6518 1990 15471 6001 1565 10583 5008 5460 13706 5565 70 1989 35178 2 363 5552 29240 14252 3379 43605 17303 1990 43532 28343 7768 35178 22363 10142 53208 22285 71 1989 24959 16064 4480 15877 13026 10524 36864 27079 1090 30311 21314 7527 24959 16064 7629 40547 29139 620 754 1337 292 900 15820 10567 72 1989 . 1811 1990 2008 970 777 1811 620 624. 15686 12468 73 1989 3859 1342 2386 3602 1137 2438 5560 13362 " 1990 2045 769 2193 3859 1342 952 25414 12628 ýý' .. ýý ý__.. , .. f irm VAR. PROF ECG. RES. L. T. F. D. loginv prorat MU lnmu no. year 167614 96000 105954 255784 11.0324 071392 0.50958 0.41183 1 1987 222862 396000 189168 280925 11.2499 0.93977 0.48008 0.39209 1988 35454 396000 28 285 313595 11.491 1.7-5406 0.46234 0.38003 1989 ... 314028 396000 37431 331571 11.367 1.16632 0.43048 0.35801 1990 120591 11880 228490 59574 10.4562 1.86245 0.24784 0.. 22142 2 1987 129665 11880 260762 161110 11.4927 2.68266 0.59092 0.46431 1988 121224 200000 67698 172322 12.3707 1.31133 0'. 64372 0.49696 1989 57182 200000 70820 268908 11.4891 0.24149 0.99294 0.68961 1990 60000 36368 58543 9.63338 0.86142 0.60749 0.47468 3 1987 .."7658 3b443 60000 41827 32131 9.56878 0.78291 0.1554 0.27425 1988 48253 60000 53458 88388 9.67055 0.96539 0.77904 0.57607 1989 60322 90000 156805 81990 9.45853 1.09196 0.33221 0.28684 1990 39348 22000 12753 0 10.5055 1.9754 0 0 4 1987 58595 22000 26921 0 10.5412 2.44197 0 0 1988 54492 22000 37850 30000 10.8491 1.03237 0.50125 0.4063 1989 64 374 22000 49575 60000 10.4751 0.74378 0.83828 0.60883 1990 26102 20000 42311 0 7.40489 1.95763 0 0 5 1987 33540 20000 49195 0 8.53228 2.55015 0 0 1988 37754 26000 49815 0 7.55904 2.28941 0 0 1989 31608 33000 54994 0 8.62209 1.9913 0 0 1990 38734 15500 108081 91261 10.1519 1.01495 0.73847 0.55301 6 1987 67585 30000 82081 106730 10.3104 1.65159 0.95226 0.66899 1988 121793 30000 88288 77395 10.4394 2.97251 0.65429 0.50337 1989 50000 80783 88079 10.4589 3.87773 0.67347 0.5149 1990 179255 . 14376 11936 8.52437 -7065 14000 -0.2146 0.42064 0.35111 7 1987 45317 14000 1825 35163 9.52069 1.723 2.22199 1.17 1988 10905 50000 28227 30022 11.1122 0.35196 0.38378 0.32482 1989 100000 62487 305960 12.6023 1.88298 1.05882 1990 -39838 -0.436 . 12540 70000 35879 78570 9.895 0.23958 0.74207 0.55508 8 1987 5626 70000 33305 76625 10.1004 0.10454 0.74174 0.55488 1988 9915 100000 3608 69377 10.1728 0.1707 0.66961 0.51259 1989 4019 73022 -72 100000 10.074 -0.0014 0.70201 0.53181 1990 58947 7500 71753 30453 9.39074 2.84372 0.38425 0.32516 9 1987 25806 7500 19797 51922 9.61787 1.59347 1'. 90211 1.06544 1988 2566 7500 18126 38771 7.34987 0.12137 1.51296 0.92146 1989 7500 20247 22117 9.76296 -30492 -1.4126 0.7971 0.58617 1990 5719 4000 4202 0 6.91075 4.30737 0 0 10 1987 9050 6000 3577 0 6.81124 7.5492 0 0 1988 8510 8000 43-08 0 7.0934 8.36772 0 0 1989 7711 12000 3143 0 7.40974 8.0592 0 0 1990 14310 30533 1960 28548 8.96738 0.73993 0.87859 0.63052 11 1987 20710 30533 2777 21241 9.09178 1.01631 0.63768 0.49328 1988 27688 30533 5408 19231 8.76358 1.2174 0.53507 0.42858 1989 37455 30533 13392 4007 8.77725 1.70758 0.09122 0.0873. 1990 '" 53548 12000 14788 13000 10.4826 0.92734 0.48529 0.39561, 12 1987 60063 12000 20771 7000 10.8836 0.85746 0.2136 0.19359 1988 83824 12000 25816 23000 11.1075 0.98421 0.60821 0.47512 1989 83568 120100 31006 23000 10.8338 0.75374 0.53481 0.42841 1990 40000 37482 0 11.0441 -59426 -0.9192 0 0 13 1987 27697 40000 14559 0 9.97315 0.45885 0 0 1988 . 1989 40000 7564 ''0 10.1711 0.34458 0 `'0 1989 20628 40000 11839 0 9.55265 0.30542 0 0 1990 24456 10800 17724 2071 9.3811 1.91943 0.07261 0.07009 14 1987 33398 17280 18428. 2360 10.1507 1.75187 0.06609 0.064 1988 38249 20736 27172 2650 9.47608 1.05548 0.05531 0.05384 1989 47156 20736 33803 2675 9.21841 1.32867 0.04905 0.04788 1990 33673 5000 22191 45702 9.66802 1.0549 0.96845 0.67725 15 1987 . 1086 60000 10852 37907 9.92662 1.05555 0.53502 0.42854 1988 270: 34 60000 1 779 38139 9.86282 0.83752 0.52404 0.42136 1989 19932 60000 13187 56361 9.52785 0.53872 0.7701 0.57103 1990 4421 8000 -2483 21232 8.12089 0.9153 3.84847 1.57866 16 1987 14224 8000 92 13000 7.24708 2.68828 1.60652 0.95802 1988 '12693 16000 4969 3000 7.79811 2.29758 0.14307 0.13372 1989 22051 16000 2964 3000 8.27995 3.27434 0.15819 0.14686 1990 18727 30000 8591 48804 10.6814 0.5125 1.26465 0.81742 17 1987 22912 30000 6736 47671 9.39955 0.37032 1.29766 0.83189 1988 27112 30000 5240 49248 9.52413 0.47737 1.3975 0.87443 1989 23822 30000 6183 17084 9.61233 0.44861 0.47216 0.38673 1990 16628 5000 14234 616 7.41095 4.72471 0.03203 0.03152 18 1987 18886 10000 14985 579 8.27232 5.11827 0.02317 0.02291 1988 27446 10000 2 554 496 7.65681 4.41566 0.01524 0.01512 1989 28864 10000 29377 345 8.35561 4.54667 0.00876 0.00872 1990 13178 5600 12639 16664 9.22779 0.76362 0.91365 0.64901 19 1987 15813 5600 14750 4279 9.16837 0.82261 0.21027 0.19084 1988 9549 9065 1730 9824 10.6496 0.46347 0.91005 0.64713 1989 15207 9065 -38 21336 9.33061 0.2588 2.36358 1.213 1990 10387 1800 3130 10675 8.55545 1.83582 2.16531 1.15225 20 1987 12933 4900 1531 15853 9.15631 1.62586 2.46509 1.24274 1988 9333 4900 3036 15063 8.15737 0.67627 1.89806 1.06404 1989 16878 4900 3560 17512 9.20059 1.3775 2.06998 1.12167 . 1990 2620 16000 -1423 42319 8.19589 0: 1764 2.90314 1.36178 21 1987 5030 16000 -1712 41320 8.1194 0.3218 2.69214 1.35896 1988 865 16000 -1366 39675 8.80072 0.05336 2.7ili5 1.31134 1989 883 200 14860 38104 8.9 0.04741 2.53015 1.26134 1990 13854 1500 15922 0 5.48894 12.6856 0 0 22 1987 15710 1500 17130 0 6.20254 18.8794 0 0 1988 12448 1500 18905 0 6.31173 14.0766 0 0 1989 9963 1500 18972 0 6.44889 10.1119 0 0 1990 8871 10030 13110 1074E 8.72291 0.36205 0.46448 0.3815 23 1987 11567 10030 16460 11165 8.10621 0.4736 0.42223 0.35223 1988 8471 10030 17950 11185 8.52695 0.4268 0.39975 0.33629 1989 8671 10030 18931 11216 7.73543 0.43539 0.38728 0.32734 1990 13139 4500 16142 7627 7.5979 1.54573 0.36949 0.31444 24 1987 15332 4500 20648 9862 7.94768 1.70755 0.39216 0.33086 1988 18057. 4500 26038 6547 7.77149 1.73406 0.21439 0.19424 1989 JL3287 4500 30260 5265 7.86902 1.22156" 0.15147 0.14104 1990 10685 5000 8005 3264 9.92064 0.65217 0.25098 0.22393 25 1987 13724 5000 7910 10060 10.0305 0.47098 0.77924 0.57619 1988 18044 5000 8097 9767 9.15631 0.41175 0.74574 0.55718 '1989 26392 7500 35956 6710 10.2401 0.67976 0.15441 0.14359 1990 3533 10000 -765 9270 8.64559 0.39034 1.00597 0.69613 26 1987 4460 10000 -981 9655. 8.8959 0.3598 1.07052 0.7278 1988 8987 10000 -877 10965 8.80762 0.62867 1.20191 0.78932 1989 11309 10000 -13 10028 9.19136 0.80261 1.00411 0.6952 1990 2712 1200 4243 16302 9.498 0.12021 2.99504 1.38505 27 1987 4191 31031 9.8293 0.25608 1.91044 6795 -1200 .. 5.75607 1988 8476 1200 4191 29686 9.64995 0.23585 5.50659 1.87281 1989 1200 4191 26262 9.52186 0.22431 4.87145 1.7701 1990 8350 10591 1444 7.40062 7.53163 0.07624 0.07347 28 1987 8350 12753 -2172 7.4922 11.716 0.10292 0.09796 1988 8350 14986 5215 8.65539 4.98461 0.22347 0.20169 1989 8768 16888 6555 8.34403 0.14934 0.2555 0.22753 1990 5000 6116 793 5.70044 1.30513 0.07134 0.06891 29 1987 5000 7097 656 5.84644 2.36055 0.05423 0.05281 1988 750C) 5502 533 6.49527 2.52398 0.04099 0.04018 1989 7500 6612 409 5.88888 1.10861 0.02898 0.02857 1990 70000 71948 23200 9.06219 0.15407 0.16344 0.15138 30 1987 70000 75724 18754 8.08425 0.41213 0.1287 0.12106 1988 70000 77394 11496 8.89179 0.91814 0.078 0.0751 1989 70000 79822 13846 8.2928 0.44181 0.09242 0.0839 1990 4500 2993 8759 8.20767 0.40364 1.16896 0.77425 31 1987 4500 5277 8732 7.90986 0.31762 0.89312 0.63022 1988 6000 4502 8393 8.19063 0.32252 0.79916 0.58733 1989 6000 5194 7488 8.18284 0.29143 0.66893 0.51218 1990 7010 3 0 7.23129 0.9798 0 0 32 1987 7010 2786 0 7.49887 1.97715 0 0 1988 7010 5792 0 7.46164 1.79723 0 0 1989 7010 10134 0 7.94272 1.9629 0 0 1990 3780 18538 0 6.57925 5.0113 0 0 33 1987 3780 20435 0 6.93245 4.4574 0 0 1988 3780 22289 0 6.53524 2.81015 0 0 1989 22401 3780 0 5.93225 4.37664 C) 0 1990 i3 0010 -264 2600 8.40447 0.17535 0.20415 0.18577 34 1987 13000 -254 2450 9.1238 0.18208 0.19222 0.17581 1988 15000 -1372 638 9.25913 -0.112 0.04682 0.04575 1989 15000 964 398 8.91825 0.15193 0.02493 0.02463 1990 4000 2904 56 7.47873 2.5948 0.00811 0.00808 35 1987 4000 4354 0 7.84424 3.44089 0 0 1988 4000 5881 C) 8.43642 3.54503 0 0 1989 4000 7696 547 9.28424 3.02193 0.04677 0.04571 1990 2000 2879 1574 7.88908 4.84484 0.32261 0.2796 36 1987 2000 7282 2502 8.20985 2.86125 0.26955 0.23867 1988 4000 10247 2080 8.18284 C). 92233 0.146 0.13627 1989 4000 11037 1688 7.82804 0.87394 0.11226 0.10639 1990 1500 3543 0 6.24998 2.48449 0 0 37 1987 1500 5445 0 6.66185 3.16204 0 0 1988 1500 7500 0 6.41673 3.39754 0 0 1989 1500 10661 0 6.72143 4.19399 0 0 1990 2614 12077 7666 7.15929 2.36736 0.521B2 0.4199 38 1987 2614 16357 9441 6.97261 2.54452 0.49765 0.4039 1988 2614 17738 9453 6.89568 3.77993 0.46448 0.3815 1989 2614 20140 8131 6.77765 3.95893 0.35734 0.30553 1990 540000 579522 11876 12.4694 2.63782 0.01061 0.01055 39 1987 257488 8712 13.9214 7.77376 350000 0.0096 0.00955 1988 550000 275957 28139 13.0865 1.79568 0.03039 0.02994. 1989 30000 64960 38813 9.51156 3.47865 0.40873 0.34269 40 1987 50000 62462 34571 9.58802 2.98752 0.3074 0.26804 1988 65000 67313 49750 9.32563 3.5693 0.376 0.31918 1989 130000 194085 112659 9.95523 2.29881 0.34762 0.29834 41 1987 L30000 207352 113811 10.6489 0.57784 0.33737 0.2907 1988 L30000. 212389 216405 10.7962 0.7944 0.63204 0.48983 1989 1200 108 435 6.46459 8.05766 0.33257 0.28711 1987 . -42 1200 1587 38B 7.46508 9.86658 6.13922 0.13034 1988' 6799 8000 4 28 7262 7.8_874 2.98024 0.86165 0.62146 49385 90000 2850 39273 11.0175 0.31129 0.42297 0.35275 43 31803 90000 4146 31207 10.4944 0.1691 0.33147 0.28629 23562 90000 -6569 34464 10.398 0.13045 0.41308 0.34577 187149 60000 143862 167924 10.668 6.8689; ' 0.82371 0.60068 44 199779 60000 196883 377989 10.6823 4. "_5: 8 1.47144 0.9048 208661 60000 243 544 479372 10.6355 3.5378 1.57925 0.9475 6.49072 427: i. 9211 1800 8932 11076 ._Y.... 1.03205 0.70905 45 1800 12297 7077. 5.87493 27.3534 0.50174 0.40662 22098 - 27732 1800 16170 5044 6.15698 46.00398 0.28069 0.2474 6647 5100 1773 4826 7.16704 0.59115 0.70217 0.5319 46 8284 5100 3168 6057 8.56674 0.92787 0.73258 0.54961 6582 5100 2421 6211 8.64927 0.73605 0.82582 0.60203 15031 12000 14200 12338 8.06589 2.28061 0.47092 0.385B9 47 14245 12000 14302 21744 9.70424 1.93546 0.82671 0.60251 11717 12000 15303 22B73 8.73569 0.52521 0.83775 0.60854 5194 750 6365 5000 7.18992 2.04364 0.70274 0.53224 48 3672 750 7430 595 7.76047 1.51245 0.07274 0.07021 2516 750 7679 998 7.39879 0.73005 0.1184 0.1119 2016 1000 2961 5291 7.82644 0.34887 1.33577 0.84834 49 2886 10000 2204 5291 7.50163 0.40192 0.43355 0.36015 3882 10000 1772 93+99 7.66996 0.51692 0.79842 0.58691 0 5925 8.85138 -983 4900 -0.1548 1.20918 0.79262 50 1352 75000 270 35,921 7.94662 0.10695 0.45315 0.37374 1743 7500 -344 3989 7.92732 0.13021 0.55743 0.44304 3069 9900 52 O 6.90274 1.7122 00 51 2285 9900 74 0 6.96224 1.83186 00 1467 6600 93 0 6.89972 1.25965 00 2170 1000 1540 10548 5.64545 3.97952 4.15276 1.63953 52 2129 1000 1701 9253 6.28227 3.09011 3.42577 1.48744 2701 iOOc7 18722 11646 6.68461 2.88337 4.05501 1.62038 1505 +8690 50583 4393 9.92701 0.02145 0.04921 0.04804 53 55374 38690 60921 45015 10.2545 0.73646 0.45191 0.37288 23328 60000 3777 19390 9.15715 1.17849 1.98323 1.09301 54 28705 6000 6993 55390 9.57199 1.42967 4.26306 1.66071 14043 3000 16306 2063 8.31874 0.99075 0.10686 0.10153 55 15593 10000 13794 6247 E3.0542 1.19225 0.26255 0.23313 6757 8000 7155 12828 7.87436 0.4982 0.84645 0.61327 56 8000 14584 17062 7.6857 0.56275 -4470 -0.3901 0.75549 12928 68000 5272 11388 8.62676 1.67944 0.94334 0.66441 57 18871 6108 13463 9.33662 2.24261 "1.04,3 0.71442 _6800 17143 5000 3552 5500 7.03743 12.4154 0.64312 0.4966 58 19923 45000 220 19000 9.48418 13.0505 0.42017 0.35078 5952 3000 13938 3000 7.05962 1.69019 0.17712 4.16307 59 7457 3000 16093 3000 7.7003 2.23401 0.15713 0.14594 2153 2000 3576 2282 7.46049 0.34441 0.40925 0.34306 60 3742 4520 5751 1847 7.70481 0.59015 0.17983 0.16537 234242 55560 164688 149347 11.1471 2.91413 0.67809 0.51765 61 175786 55560 188945 139141 11.5459 1.71484 0.56891 0.45038 250000 5929 286254 10.965B -0.0382 1.11849 0.7507 62 45253 250000 16024 251588 11.663 0.14496 0.94573 0.66564 47425 12000 74605 0 9.73572 7.14777 00 63 97762 12000 103271 0 9.90793 '5.26319 00 35310 9800 -7745 0 8.78661 1.64807 00 64 8362 9800 -1900 10972 8.05833 0.48644 1.38886 0.87082 28163 33000 22800 36490 B. 88682 0.88794 0.65394 4.50316 65 3 4000 22452 22692 8.46168 0.78511 0.40197 0.33788 1990 11069 30600 164 71920 7.68386 1.27374. 2.3378 1.20531 66 1989 1143 38250 1869 69779 9.04794 2.44705 1.7393 1.0077 1990 12278 7800 5493 13428 8.56751 2.24315 1.01016 0.69821 67 1989 6263 5994 8.52853 20796 7800 . 82837 0.42622 0.35503 1990 9634 904 2500 965 8.52238 1.5468 0.8382 0.24984 68 1989 6996 900 4163 701 7.93308 0.7951 0.13846 0.12967 1990 11579 2343 7600 0 0.33591 2.32514 0 0 69 1989 8+. 2=.G^ 1 343 7646 0 8.6052 1.39788 0 0 1990 26302 16000 4597 3591 8.12534 1.65306 0.17435 0.16071 70 1989 30923 16000 12984 3445 9.22444 2.28745 0.11886 0111231 1990 9785 4000 6419 29942 9.26141 3.23301 2.87379 1.35423 71 1989 11408 40010 7023 22799 8.93971 1.21577 2.06831 1.12113 1990 5253 8000 382 0 6.80239 4.73516 0 0 72 1989 3218 6000 1883 0 6.43615 2.56131 0 0 1990 15169 14556 7.79893 -7602 12000 -2.9815 0.53576 0.42902 73 1989 12786 1 000 15144 13730 6.85857 4.81549 0.50582 0.40934 1990 xýýý -fi ý , ý, -ýyr _ _' "-ý3 rý ,; ,t*Iýi r25. tý2y_ý ý ; 4'ý .. ... Ir ii-.. fý n ... ., c . _. .. ý .ý .ý,... ý.. . ..,.. . Clothing sector firm ACC. D. no. year K ACC. D. DEFT;. <(-1) t-1 INV. V. ADD. LAB. C. VAR. PR. 1 1987'137643 89862 12023 147894 89185 1095 270538 42394 228144 1988 181628 105425 16492 13764 3 89B62 44914 221352 50535 170817 1989 192682 123471 19342 18168 105425 12-350 168198 59709 108489 1990 194767 129555 18746 192682 123471 14747 219299 61452 157847 2 1987 103100 81979 93Ö0 91735 69012 7698 176256 156304 19952 1988 103554 84957 e120 103100 81979 5604 187199 172692 14507 1989 118593 92694 10112 103554 84957.17414 167232 147767 19465 1990 124236 96463 7992 118593 92694 9866 157003 161104 -4101 3 1987 33967 10731 :3,614 28219 8866 7497 67966 15434 52534 1988 35202 12703 4451 33967 10731 3714 69369 19133 49236 1989 40003 15356 3 724 35202 12703 5872 70938 23181 47757 47130 18220 5310 . 1990 40003 15356 9573 88295 26591 61704 4 1987 4325 2510 670 3882 1874 477 50931 4252 46679 1988 4530 3100 677 4 325 2510 2292 41169 4715 36454 1989 5351 3874 686 4530 3100 73 48700 5745 42955 1990 6683 4729 27961 5351 3874 28438 61334 . 5738 55596 5 1987 17724 8869 4679 15696 7570 5408 34 333 22742 11591 1988 21571 10349 4037 17724 8869 6404 42650 25385 17265 1989 22697 11451 4403 21571 10349 4427 45425 29454 15971 1990 25976 12758 4472 22697 11451 6444 53243 30242 23001 6 1987 22794 13690 2793 20186 11357 3068 37607 25457 12150 1988 25801 16012 3106 22794 13690 3791 43224 27739 15485 1989 28375 18013 3257 25801 16012 3830 43120 29763 13357 1990 30043 20491 3371 28375 18013 2521 45526 30931 14595 7 1987 9994 3797 1104 10754 3466 13 19546 12201 7345 1988 10929 5049 1081 9994 3797 764 17919 11351 6568 1989 1 182 5887 1384 10929 5049 1799 19020 12403 6617 1990 11914 6673 1436 12182 5887 382 21259 13,378 7881 8 1987 11568 7009 1986 10809 5142 878 18135 6943 11192 1988 12630 7645 1243 11568 7009 1669 21264 9277 119? 7 1989 13137 8531 1136 12630 7645 757 29062 11951 17111 1990 14535 8527 1599 13137 8531 3001 25388 15420 9968 1987 20348 9896 1028 18588 9 8920 1812 33921 34908 -987 1988 20692 10842 1303 20348 9896 701 36924 36747 177 1989 20621 10670 1347 20692 10842 1448 42583 37186 5 397 1990 22957 11712 1557 20621 10670 2851 47946 37462 10484 10 1987 36076 19802 5890 27446 14526 9244 25265 12099 13166 1988 49156 26124 6677 36076 19802 13435 37176 15184 21992 1989 56449 31716 6631 49156 26124 8332 48754 17928 30826 1990 63029 37503 7961 56449 31716 8754 44418 19103 25315 11 1987 12599 6464 2656 12610 5841 2022 16533 11085 5448 1988 12546 7165 2613 12599 6464 1859 18523 9090 9433 1989 12795 7970 2740 12546 7165 2184 18841 9321 9520 1990 13000 8132 2920 12795 7970 2963 29764 10813 18951 12 1987 13720 8675 1512 11252 7678 2983 32122 27515- 4607 1988 17768 9828 1471 13720 8675 4366 34759 29599 5160 1989 18805 11170 1706 17768 9828 1401 30694 31908 6786 1990 19671 12625 1008 18805 11170 419 41549 33621 7928 13 1987 8841 6095 978 8369 5163 498 21221 12228 8993 1988 11789 7943 1043 8841 6095 2143 22018 13350 8668 7727 1030 11789 1989 12231 7943 '1688 22976 146631`° ' 8313 1990 13546 8557 1400 12231 7727 -1885 26135 16190 "° 9945 14 1987 14447 6489 886 14655 6009 198.11669 1941" 9728 1988 14529' 7088 694 14447 6489 377 10171 2245-, *, 7926 14646 7514 572 14529 -z. 1989 7088 263 11460 =3457., 8003 1990 15145 7836 430 14646 7514 607 8847 3723 5124 15 1987 8874 3933 1112 7846 2965 1172 25505 19252 6253 , 1988 9709 4845 1164 8874 3933 1087 30037 21948 8089 1989 12224 5619 905 9709 4845 2646 34472 25399 12224 9073 1990 0877 6465 1222 5619 9029 37722 26329 11393 16 1987 26455 13453' 2908 23292 10896 3514 20366 14428 5938 1988 31838 16255 3234 26455 13453 5815 22157 15465 6692 1989 34289 18666 2989 31838 16255 3029 26665 17856 8809 1990 35695 19725 1992 34289 18666 2339 26046 19051 6995 17 1967 9030 4540 407, 9376 4212 529 15236 4314 10922 1988 14396 5670 1000 9830 4540 4436 16335 5093 11242 1989 20014 6710 1106 14396 5670 5684 9964 5733 4231 1990 21834 7727 1148 20014 6710 1951 13937 5925 8012 18 1987 14984 10869 1071 14284 10179 1081 28210 12173 16037 1988 21815 11885 1214 14984 10869 7029 33816 14107 19629 1989 26686 13049 1337 21815 11885 5044 41486 16889 24597 1990 29847 14566 1757 26686 13049 3401 45573 18985 26588 19 1987 14677 10094 2558 12449 7569 2261 10074 4533 5541 1988 16064 11916 2145 14677 10094 1710 9729 4341 5388 1989 15968 12576 978 16064 11916 222 5470 4187 1283 1990 17451 13642 1226 15968 12576 1643 7077 3618 3459 20 1987 14319 7472 2309 12990 6668 2829 21843. 16999 4844 1988 15192 8364 2569 14319 7472 2550 21487 16323 5164 1989 15036 6949 3273 15192 8364 2532 20844 15402 5442 1990 14528 8680 2121 15036 8949 1882 17928 13767 4161 21 1987 12650 8772 1464 12389 7415 36B 13497 6736 6761 1988 5546 1062 1205 12650 8772 1811 15247 7655 7592 1989 6659 2031 1145 5546 1062 1289 22029 9067 12962 1990 10826 3533 1785 6659 2031 4450 24060 10988 13072 22 1987 21489 11789 1186 21509 11267 644 19714 16515 3199 1988 21564 13020 1423 21489 11789 267 24394 18661 5733 1989 25857 14145 1214 21564 13020 4382 23189 20611 2578 1990 26514 15149 1148 25857 14145 801 23843 21220 2623 23 1987 9667 7849 1002 9445 6954 329 15425 13176 2249 1988 11739 8377 697 9667 7849 2241 15417 14442 975 1989 14371 8981 942 11739 8377 2970 16824 15448 1376 1990 15793 9836 1131 14371 8981 1698 18459 16762 1697 24 1987 7173 4859 593 6943 4344 308 17537 15101 2436 1988 82.64 5448 643 7173 4859 1145 18280 15579 2701 1989 8352 5994 719 8264 5448 261 22423 16754 3669 1990 8555 6453 621 8352 5994 365 24183 20516 3667 25 1987 1975 1173 306 1652 993 449 6398 2472 3926 1988 2637 1617 487 1975 1173 705 7818 3059 4759 1995 856 1989 ----3279 2637 1617 1120 10362 3981 6381 1990 4333 2229 1227 3279 1995 2047 11790 4585 7205 26 1987 15290 6048 1296 15317 5102 323 13552 9865 3687 1988 16166 7006 1253 15290 6048 1171 149.25 10888 4037 1989 16932 8040 1294 16166 7006 1026 16940. 13204 3736 1990 17179 8956 1260 16932 8040 591 16460 13039 3421 27 1987 8754 4668 1862 7847 3357 1458 25254 9685 15569 10509 5277 891 8754 4668 1988 2037 19108 -13778. 5330 1989 13762 6303 1427 10509 5277 3654 18593 14660 3733 1990 16771 6705 936 13762 6303 3543 18283 14896 3387 28 1987 17144 11934 1535 15895 10479 1329 28247 20201 8046 1988 18394 13206 1388 17144 11934 1366 25358 20451 4907 1989 19338 13419 1126 '18394 13206 1857 25376 21150 4226 1990 20085 14084 1126 19338 13419 1208 26133 21150 4983 29 1987 5405 1466 306 5221 1191 215 6457 3037x,,, -3420 - 1988 5697 1746 361 5405 1466 373 8228 3563 4665 1989 5871 2056 379 5697 1746 243 8719 4012 4707 1990 7785 2374 440 5871 2056 2036 8714 4387 4327 Z,0 1987 23470 17178 1842 21385 15623 2372 60047 19003 41044 1988 24930 17590 2203 23470 17178 3251 68463 14994 53469 1989 24323 18119 2958 24930 17590 1822 46955 15375 31580 31 1987 27891 14421 2801 24233 12169 4207 29817 8345 21472 1988 32075 15574 2446 27891 14421 5477 20890 10143 10747 1989 32214 15565 2769 32075 15574 2917 25741. 11404 14337 32 1987 8694 5348 3944 7738 3846 3398 22991 5294 17705 1988 9692 6839 3775 8694 5348 3282 30616 6458 24158 1989 11700 7746 3589 9692 6839 4690 37039 7858 29181 33 1987 7084 4986 1102 6367 4233 1066 17460 3861 13599 1988 7567 5637 1078 7084 4986 910 20846 6762 14084 1989 7907 6129 1025 7567 5637 878 22543 6087 16456 34 1987 4632 1817 677 4306 1304 490 6250 4273 1977 1988 5246 2070 327 4632 1817 688 5599 4235 1364 1989 5281 2155 335 5246 2070 285 6519 4695 1824 1988 21174 2224 1323 7910' 1172 13535 25434 14912 10522 . 35 1989 30044 5284 3331 21174 2224 9141 29664 16210 13454 1990 35462 8653 5238 30044 5284 7287 2877 18063 -15186 36 1988 10308 4500 1759 6798 3014 3783 15304 6054 9250 1989 11387 5913 1794 10308 4500 1460 15925 7219 8706 1990 12200 6972 1681 11387 5910 1435 18584 7941 10643 37 1988 6702 2119 680 6751 1739 251 7760 1716 6044 1989 7045 2430 521 6702 2119 553 7063 1377 5686 1990 10176 2723 613 7045 2430 3453 7962 1408 6574 38 1988 8382 2891 3635 3909 2338 7555 16147 7955 8192 1989 9975 3735 3948 8382 2891 4697 21831 11272 10559 1990 12924 4736 4405 9975 3735 6353 24061 13277 10784' ti 1988 2048 1287 496 1776 1044 525 12632 3521 9111 1989 2129 1447 435 2048 1287 356 11718 2344 9374 1990 3235 1610 400 2129 1447 1343 16624 2048 14576 40 1988 7814 3590 518 5578 3190 2354 18250 15113 3137 1989 9532 4259 855 7814 3590 1904 20942 17310 3632 1990 10420 5302 1225 9532 4259 1070 21701 17643 4058 41 1988 25041 19003 1210 23943 18345 1650 26430 23932 2498 1989 25746 19392 1278 25041 19003 1594 25324 24605 719 1990 . 26567 20313 1555 25746 19392 1455 25239 24586 653 42 1987 13706 7341 2417 12020 5336 2098 19119 9252 9867 1988 15206 8899 1782 13706 7341 1724 17502 10572 6930 43 1987 8465 5093 627 7013 4716 1702 17158 5418 11740 1988 10109 5690 820 8465 5093 1867 20068 6010 14058 44 1987 317 126 243 265 80 249 4034 6389 -2355 1988 371 162 393 317 126 411 8658 4751 3907 45 1987 3271 1384 838 2222 959 1462 4207 2872 1335 1988 3582 2166 1045 3271 1384 574 3988 3763 225 46 1987 3870 1859 441 2990 1614 1076 5697 7000 -1303 1988 5273 1990 1200 3870 1859 2472 11196 6775 4421 47 1989 10731 3594 4075 8507 2835 5540 83356 56454 26902 1990 15745 5883 6111 10731 3594 8836 83196 63063 20133 48 1989 27186 16440 2269 24543 14207 2679 50477 39364 11113 1990 31054 19205 2915 27186 16440 4018 46818 31890 14928 49 1989 31341 16812 2595 31278 15126 972 26345 19672 '16673 33277 17831 3791+ 31341 16812 4708 24814 17270 ;7544 1990 ,: std 1989 20033 601 2462 12226 3885 8141 12462 `7577 `4885 1990 243 24 8972 33 27 20033 6013 4659 15275 6352 6923 51 1989 20956 13209 2322 19075 11213 2207 21409 11472 9937 1990 2: 3627 15583 2566 20956 13209 2863 24043 12653 11390 52 1989 11849 5960 973 10814 5435 1463 20722 15451 5271 1990 12586 6762 1390 11849 5980 1345 9951 16031 -6080 53 1989 54154 21507 2968 42422 18543 11736 29915 21501 8414 1990 66954 24649 3154 54154 21507 1281 29064 23898 5166 q ;ý .ýý `jam ý \_ýr. _r . t, `h. f irm EQ. RES. L. 'r. F. D. loginv prorat mu lnmu no. year 745Ü0 73702 10043 6.99851 3.661457 0.067766 0.065568 1 1987 74500 59188 7986 10.7125 3.402622 0.059736 0.05802 1988 7450() 61394 5852 9.421411 1.341087 0.043063 0.042162 1989 74500 66216 4480 9.598795 2.161971 0.031837 0.031341 1990 28000 67283 7315 8.948716 0.827314 0.076771 0.073967 2 1987 28000 82112 8552 8.631236 0.653734 0.077666 0.07479.8 1988 28000 91652 9539 9.765c: 3 0.98595 0.079723 0.076704 1989 28000 109118 21937 9.19685 -0.15011 0.159986 0.148408 1990' 65000 43176 6163 8.922258 2.557654 0.056972 0.055408 3 1987 651000 59965 5094 8.219865 2.016783 0.040763 0.039954 1988 71500 69114 11761 8.677951 1.99948 0.06364 *0.0803 26 1989 71500 69805 17702 9.166702 2.37322 0.125275 0.118028 1990 200 21220 0 6.167516 21.90319 0 0 4 1987 200 5477 0 5.676754 19.11641 0 0 1988 200 8323 0 6.597146 28.29573 0 0 1989 36035 144828 0 10.25548 35.68221 0 0 1990 6250 14952 6000 8.595635 1.343983 0.282 992 0.249195 5 1987 6250 15586 6000 8.764678 1.855734 0.274776 0.24277 1988 6250 16181 6000 8.395477 1.340618 0.267487 0.237036 1989 6250 17064 6000 8.770905 1.938819 0.. 2 57356 0.229011 1990 6920 15890 10771 8.028781 1.296625 0.472205 0.386761 6 1987 13840 11209 10128 8.240385 1.618888 0.404328 0.339559 1988 13940 1248 7064 8.25062 1.285327 0.268369 0.237732 1989 13840 13999 7006 7.832411 1.335209 0.251661 0.224472 1990 10265 1924 10749 2.564949 0.949583 0.881861 0.632261 7 1987 10265 1984 7146 6.638568 1.008764 0.583395 0.459571 1988 10265 2137 11029 7.494986 1.060052 0.889292 0.636202 1989 2293 10265 9767 5.945421 1.186791 0'. 777751 0.575349 1990 1000 10147 11524 6.777647 -1.860819 1.033821 0,. 709916 8 1987 13028 9927 7.41998 2.502527 0.707656 0.535122 1988 1000 . 50000 11786 10299 6.62 9363 3.2 33355 0.613547 0.478435 1989 5000 15714 7711 8.006701 2.051506 0.37226 0.316459 1990 15000 496 2718 7.502186 -0.09619 0.1754 0.161609 9 1987 15000 -2133 2441 6.552508 0.016118 0.18971 0.17371 1988 15000 0 1704 7.277939 0.51613 0.1136 0.107598 1989 15000 -769 895 7.955425 0.998732 0.062891 0.060992 1990 3600 2014 20204 9.13173 0.960154 3.59886 1.525808 10 1987 3600 4317 29336 9.505619 1.286199 3.705444 1.54872 1988 E600 10706 25375 9.027859 1.260749 1.314358 0.839132 1989 8600 21867 27359 9.077266 0.970264 0.897986 0.640794 1990 7016 9365 6690 7.611842 0.758337 0.4084 0.342454 11 1987 7016 -527 9500 7.527794 1.463434 1.464016 0.901793 1988 7016 468 9500 7.688913 1.666545 1.269375 0.819504 1989 7016 2 404 19500 7.993958 3.723261 2.070464 1.121698 1990 5000 4746 5260 8.000685 1.214544 0.539709 0.431593 12 1987 5483 5117 8.381603 0.973478 4.488124 4.397516 1968 5000 , 5000 6340 5827 7.244942 0.805075 0.513845 0.414653 1989 5000 7743 5027 6.037871 0.9843 36 0.394491 0.33253 1990 50C)0 5195 6333 6.2106 2.626575 0.621187 '0.483159 13 1987 4990 7125 5825 7.669962 3.004389; 0.480809 . 0.392589 1988 4990 9584 7183 7.4313 2.036065 0.492864 0.400696 1989 5385 7.541683 2.093125 499 15012 . -0: 269223 '0.239405' -' 1990 1000 4286 31968 5.88267 1.060127 6.047673 1.95: 698 14 1987 1000 6601 26103 5.93 245 0.947955 3.434153 1.489337 1988 1000: 8640 220717 5; 572154 1.013129 2.149 X66 1.147106 1989 1 000 8792 16536 6.408529 0.681062 1.688725 00.989067 1990 2650 3338 600 7.066467 1.207061 0.1002 0.095492 15 1987 2650 4199 200 6.991177 1.558181 0.029201 0.028783 1988 60900 2180 0 7.880804 1.757116 0 1989 8000 2939 5000 9.108197 1.635136 00.45708 0.376435 1990 24C)00 7565 677 8.16451 0.451345 0.021448 0.021221 16 1987 24000 8600 3384 8.668196 0.489873 0.103804 0.09876. 1988 24000) 10342 4484 8.015988 0.532499 0.131)569 C). 12271 1989 24000 11579 3912 7.757479 0.424436 0.1U995"ß 0.104317 1990 6200 1788 11000 6.270988 1.992808 1.101322 0.742566 17 1987 1.4000 622 1 000 8.397508 2.022673 0.820681 0.599211 1988 140000 4742 14000 8.64541 0.456742 0.746985 Ci. 557892 1989 14000 7841 14000 7.576097 0.570883 0.640996 0.495304 1990 1500 32468 3400 6.985642 3.680947 0.100094 0.095396 18 1987 1500 41450 2664 8.8578 4.540108 0.062026 0.060178 1988 1500 51381 1866 8.525955 2.33333, C). 035287 0.034678 1989 1500 64351 541 8.13185 1.848228 0.008216 0.008182 1990 3500 4396 1800 7.723562 1.069838 0.227964 0.2053 57 19 1987 35007 5452 6250 7.444249 1.118962 0.698168 0.5955 1988 35000 7211 6750 5.402677 0.291361 0.630193 0.488699 1989 3500 7981 5965 7.404279 0.966682 0.519554 0.418417 1990 2C>- 3723 4665 7.947679 0.721366 0.815132 0.. 596158 20 1987 20OC 4224 3771 7.843849 0.717833 0.60588 0.473672 1988 6000 4489 3297 7.836765 0.750772 0.314329 0.273326 1989 6000 5899 2667 7.54009 0.648012 0.224136 0.202236 1990 2692 9794 1388 5.908083 1.280722 0.111165 0.105409 21 1987 5000 0 1198 7.501634 1.863315 0.2396 0.214789 1988 SOC'C0 2051 975 7.161622 2.723012 0.138278 0.129517 1989 50c? c1 6781 710 8.400659 2.677549 0.060267 0.05852 1990 16750 6517 401 6.467699 0.294292 0.017235 0.017088 22 1987 16750 6956 286 5.587249 0.562533 0.012064 0.011992 1988 167500 7202 1462 8.385261 0.284227 0.061039 0.059248 1989 16750 6123 2374 6.685661 0.212303 0.095445 0.091161 1990 23 30000 2204 2194 5.796058 0.850678 0.421599 0.351782 1987 4500 1054 3000 7.714677 0.510445 0.540151 0.431881 1988 5500 1402 4500 7.996317,0.385535 0.651985 0.501978 1989 E30C00 1522 5100 7.437206 0.298457 0.535602 0.428922 1990 150)0) 1190 1220 5.7301 0.883122 0.460377 0.378695 24 1987 1500 1352 802 7.04316 1.110962 0.281206 0.247802 1988 2500 677 2000 5.56452 1.227321 0.629525 0.488288 1989 2500 1257 1775 5.899897 1.474198 0.472451 0.386929 1990 1200 683 0 6.107023 5.613251 0 0 25 1987 2000 376 0 6.558198 5.647798 0 0 1988 2000 1185 0 7.021084 5.892936 0 0 1989 2500 10 315 720 7.624131 5.3193 39 4.056184 0.054663 1990 60000 0 5.777652 0.340083 0 26 1987 -802 .0 6000 -658 0 7.065613 0.415748 0 0 1988 bCU44 585 340 6.933423 0.3 841,98 0.051632 0.050344 1989 6000 1789 3739 6.381816 0.364706 0.480036 0.392066 1990 3.266311 673ä 4671 8191 7.284821 0.718257 0.54131 27 1987 6733 10742 7778 7.619233 1.236567 0.445093 0.368174 1988 6733 10134 38090-8.203576,0.685393 2.258256 1.181192 1989 6733 10145 36159 8.172729; 0.429644 2. ' 142375 1.144979 1990 t. 982 1215 6526 7.192182 1.427068 0.296879 0.259961 28 1987 9875 13181 5325 7.219642 0.891154 0.230959 0.207794 1988 9872 14270 7790 7.526718 0.767501 0.322674 0.279656 1989 9875 14547 6145 7.096721 0.801273 0.251617 0.224437 1990 3000 450 6116 5.370638 0.799596 1.772754 1.019841 29 1987 3000 610 7360 5.921578 1.127206 2.038781 1.111457 1988 5,000 885 7729 5.493061 1.122226 1.313339 0.838692 1989 4866 7.618742 1.07518 0.802705 0.589289 , (i00 1062 1990 16300 32617 205 7.771489 6.711599 0.004191 0.004182 30 1987 16300 60336 182 8.086718 8.088186 0.002375 0.002372 1968 16300 74210 59072 7.50769 4.052838 0.652657 0.502384 1989 ;5)o0 27,628 10009 8.344505 1.676991 0.337822 0.291043 31 1987 6005 30540 9930 8.6083,13.0.759377 0.271757 0.240399 1988 6000 30727 10135 7.978311 0.818448 0.275955 0.243695 1989 5000 13258 3000 8.130942 4.286202 0.164312 '0.15213 32 1987 5000 19798 0 8.096208 6.871837 00 1988 5000 27663 30000 8.453188 9.634775 0.91847 0.651528 1989 2000 13821 134 6.971669 6.004296 0.00847 0.008434 33 1987 2000 19678 0 6.813445 6.389374 00 1988 2000 24775 20000 6.771936 8.031749 6.746965 0.55788 1989 2000 337 441 6.194405 0.620505 0.188703 0.172863 34 1987 2000- 590 952 6.533789 0.461184 0.367568 0.313034 1988 4000 595 1233 5.652489 0.540988 0.268335 0.237705 1989 10000 1050 21089 9.513034 1.486295 1.908507 1.06764 35 1988 10000 2193 22661 9.120525 0.668783 1.858525 1.050306 1989 10000' 3078 23869 8.893847 -0.58141 1.825126 1.038553 1990 2400 5342 5780 8.238273+ 2.326636 0.746577 0.557658 36 1988 24.00 7493 5890 7.286192 1.412002 0.59537 0.467106 1989 2400 8096 8646 7.26892 1.843096 0.823742 0.600891 1990 7800 '4890 7800 5.525453 1.14776 0.614657 0.479123 37 1988 6.434382 7800 6538 7800 6.315358 1.168692 0.544009 1989 7800 8050 7800 8.146999 1.350351 0.492114 0.400194 1990 4980 797 5650 8.929965 4.963083 0.978016 0.682094 38 1988 4980 1694 13028 8.454679 1.811401 1.952053 1.082501 1989 4980 3379 13113 8.756682 1.638264 1.568728 0.943411' 1990 5000 658 0 6.263398 11.84657 00 39 1988 9000 855 0 5.874931 11.60335 00 1989 5000 3573 0 7.202661 20.26015 00 1990 1800 1101 2785 7.763871 1.250311 0.960014 0.672952 40 1988 18000 1331 3646 7.551712 0.809963 1.164484 0.772182 1989 1600 1677 4266 6.975414 0.72953 1.22692 0.800619 1990 3200 5452 6250 7.408531 0.424715 0.722376 0.543705 41 1988 3200 5621 1303 7.374002 0.112171 0.147716 0.137774 1989 0: :3200 5663 963 7.282761 097421 0.108654 0'. 103147 1990 1.390907 5000 4852 2649 7.64874 0.268879 0.2238134 42 1987 5000 7601 2432 7.452402 1.036269 0.193001 0.176472 1988 4.815669 10500 13080 2000 7.439559 0.084618 0.081412 43 1987 0 7.532088 3.968019 10504 18782 00 1988 3000 1815 1500 5.517453 -11.9941-0.311526 0.271192 44 1987 1857 0 6.018593 19.46919 '` 0.. 0 1988 3000 " ,. 426 13850 7.28'7561 0.995927 1.999711 1.096516 45 6500 . 1987 6.352629 0.113488 6500 501 14059 2.008142 1.101323 1988 8000 123 - 675 6.981006. -0.89223,0: 083097.0.079825 46 1987 2.092407%0.333464 8000 2196 3400 7.812783 0.228778 1988 8.61975 4.46777710,041695 28000 41456 2896 0.04085 47 1989 9.08659 2.674124=. '0.03579 28000 46601 2670 0.035165 1990 4950 42112 0 7.893199 1.012796 0 0 48 1989 4950 6010 0 8.29854 1.316872 0 0 1990 7688 2428 14984 6.879356 0.389169 1.481218 0.90875 49 1989 9610 1396 20395 8.457018 0.492215 1.85308 1.048399 1990 2500 4147 3155 9.004668 0.551683 0.47465 0.388421 50 1989 500 4760 13658 8.446556 0.468096 1.881267 1.05823 199f? 4500 3006 4500 7.699339 1.190599 0.59952 0.469704 51 1989 4500 6081 4500 7.959625 1.393731 0.425291 0.354376 1990 45000 968 2408 7.288244 0.92307 0.44038 0.364907 52 1989 4500 1246 2129 7.204149 -0.98204 0.370519 0.315189_ 1990 46010 6960 20465 9.370416 0.331917 1.740221 1.008039 53 1969 4800 7889 33729 9.458138 0.150003 2.658129 1.296952 1990 E#ý ý Rr ,. _t, 181 CHAPTER 6 CONTROLLING GROUPS, MARKET POWER, AND THE COST OF CAPITAL IN A NON- SECURITIZED FINANCIAL SYSTEM: AN ALTERNATIVE INTERPRETATION OF THE FIRMS' INVESTMENT DECISION *I am very grateful to Norman Ireland for his invaluable advice. I Keith Cowling, am also indebted to Giovanni Amisano and Flavio Rovida for their suggestions and comments at different stages of my interesting Fitzroy, work. A very conversation with Felix David Cobham, and other members of the staff of the Department of Economics at the University of St. Andrew (Scotland) in April 1992 interest for issues stimulated my and curiosity some of the analyzed in this paper. Obviously none of the above-mentioned people are that responsible for any mistakes might be found or for the views expressed here. 182 CONTROLLING GROUPS, MARKET POWER, AND THE COST OF CAPITAL IN A NON- SECURITIZED FINANCIAL SYSTEM: AN ALTERNATIVE INTERPRETATION OF THE FIRMS' INVESTMENT DECISION 1. Introduction The contents of this chapter might be regarded as a digression starting from the theoretical analysis of the investment decision of the firm of the previous chapter. The present analysis is, moreover, largely based on the same type of literature and very much concerned in with the same problems analyzed the previous chapter. For this introductory reason, the present section does not contain the usual brief survey on the relevant literature. The purpose of this chapter is to assess and analyze the implications - for the investment decision of the firm - of a few results of the industrial economics In and finance literature. addition, some emphasis will be placed on "anomalies" institutional a few financial and features that might "non-securitized" financial characterize the systems. For several the be deliberately reasons, many parts of analysis will conducted in a qualitative way. First of all, it might useful to attempt to than detail, -describe, in more usual the multiplicity of causal links that characterize the interaction between finance and firm's investment decision. This attempt might not be easily done by the "stylized facts" usual description of through simplifying functional links, because such an approach might cause some loss of information, even if it has the big merit and advantage of yielding blgebraically tractable , models focusing on a few important , few important mechanisms. These mechanisms might indeed not exhaust links all of the possible causal existing between some structural 183 features (such as market power), the cost of capital, and a few financial decisions. Obviously, it will not be possible here to few important perform a complete and exhaustive analysis, but a theoretical contributions of industrial economics and finance - suggesting the simultaneity of the financial and investment decision to be taken into - need, nevertheless, explicitly account, and (since they often result in causal links of opposite sign) not all by of them at the same time, and means of a qualitative analysis. Finally, as in the previous chapter, another goal of the present framework analysis is to provide a where the simultaneity of financial and investment decisions is explicitly taken into account. Section 2 introduces the main assumptions of the present 3 discusses few links between analysis. Section a possible causal the real decisions of the firm and the cost of capital. Section 4 formalizes the more general decision problem of the firm, which includes a set of "financial decisions" - discussed in section 5- "real decisions", discussed in 6. 7 and a set of section Section contains a few conclusive comments. 2. The main assumptions of the analysis. In the standard models of finance it is assumed that the market for shares is associated with a market for firms' control, acting as incentives a mechanism of and control on managers' behaviour: a inefficiently (in management acting would cause an efficient market) the firm's share price to fall, increasing the probability of a hostile take-over, and putting at risk the job of the management itself. This theoretical assumption requires a situation of highly 184 dispersed shares ownership, such as the one of the British and American "securitized" financial systems. The fact that the "managerial revolution" approach took place iný the Anglo-Saxon institutional contexts, where the theoretical implications of the separation between ownership and control is an important issue, could be pertinent here. While in the U. K. and in the U. S. hostile takeovers constitute a very frequent phenomenon, they are relatively rare in Germany (not more than four or five per year, as De Felice [1992] et al. [1988] and Mayer points out) and almost completely absent in Italy and in most of the "non-securitized" financial sectors14. In non-securitized financial systems, a very little dispersion-of firms' ownership is associated with the persistence of the old traditional family clan, who typically leads its financial holdings and some of the most important firms of the group (at effectively and efficiently least from their point of view! ). Another feature of these-institutional contexts is the fact that stock markets are rarely a source of financial funds, compared to intermediated credit. In particular, the transactions in the stock the larger market mainly concern, financial groupsl5. (and In this section the following subsections), it is assumed in that the management act the interest of the controlling group of individual shareholders, while the shareholder herself is regarded funds. as an external supplier of Further, it will be assumed that is the market for shares not associated with a market for control, financial while "informal" markets exist for the firms whose shares 14 For a discussion on the main characteristics of "securitized" and "non-securitized" financial systems see, for example, Gardener [1991] and Gardener and Molyneux [1990]. 29 larger financial 15 In 1987, the holdings had issued 94% of in the Italian shares traded stock market, according to the data Buzzacchi provided by Brioschi, and Colombo [1990]. 185 are not traded in the stock market. All these assumptions are meant to describe a "non-securitized" financial sector, where,, in addition, the management is directed by the group in control. In Italy, for example, such a situation applies not only to small or average sized firms, but also to most of the giant firms, issuing shares in the stock market. The main assumptions of the present analysis are the following: i) The goods market is assumed to be imperfectly competitive, although perfect competition can be a limit case of the framework of in analysis employed this paper; ii) The management is composed of members of the controlling firm in group of the and acts the interest of the controlling group of the firm; iii) The stock market is not associated with a market for firm's control; in other words, takeovers, mergers and any transaction having as an object the firm's control are performed by means of private negotiations among the managements of the different firms; iv) On the basis of assumptions ii) and iii), the controlling finance group of shareholders the physical capital with their own (i. by by wealth e. subscribing shares), or retaining profits, or by funds, be raising external which can debts or shares allocated to In non-controlling shareholders. other words, the following balance hold: sheet constraint must always k=E+R+B (1 the where k= value of physical capital (net of the accumulated depreciation); E= Psub N, i. e. equities defined at their "subscription" price; R= reserves; B= financial debt. The 186 is financial firm. expression "E +R+ B" then the capital of the Following the finance literature, its cost is the discount factor for the future streams of income. v) The decisions of the firm are assumed to be taken in but known continuous time, are made to outsiders only at discrete intervals (i. e. when the accounting data are available to outsiders). For what concerns point i), we define the profit function in the same way as in the previous chapter: u°(k, w I Pi) (2 where k is the physical capital, w, the cost of variable inputs, pi is the profit margin, determined by the demand degree differentiation, the elasticity, the of product and all other the the firm. aspects affecting market power of For assumption ii) it might be worth mentioning a recent by Blanchard [1993], the behaviour empirical study et al. on of that that managers with cash windfalls, shows evidence contrasts it "... with the assumption of perfect capital markets models, while be to fit the information needs to stretched considerably asymmetric in the interest model in which managers act of shareholders, (... ) the behaviour". and supports agency model of managerial However, In those institutional contexts without the structural and institutional phenomenon of dispersion of firms' ownership, the the lack persistence of controlling groups and of a market for firms' control might take the appearance of a self-sustaining if that the insiders better informed mechanism, one assumes are on firm's investments the quality of the than the outsider potential fact, buyers of the shares. In the assumption of highly dispersed 187 the "radical to share ownership, may raise economics" objection as have decided to why should the original controlling groups give up their control and switch to a configuration of highly dispersed higher dispersion (and ownership16. The relatively share ownership higher frequency of hostile take-overs) which characterizes the "securitized" financial systems (i. e. mainly the U. S. and the U. K. ) find justification. However, the can certainly a reasonable economic "radical" be least in above-mentioned objection could accepted at "weak terms", assuming, in other words, that the economic conditions that would induce the original controlling groups to gradually disperse their controlling shares might not have occurred for all institutional contexts. Buying a non-majority share is equivalent to performing a financial portfolio investment, since the shareholders who do not belong to the controlling group cannot interfere neither on the the firm, the decision the strategic choices of nor on concerning distribution of dividends, and their motivation lies entirely in the (dividends their The remuneration plus capital gains) of shares. in this framework stock market only provides a constraint on the behaviour of the managers (operating through a mechanism of incentives), compelling them to distribute dividends in order to the remunerate the shareholders at market rate of returnl7. 16 For a discussion, see Cowling [1982], ch. 4 17 The individual's decision on whether to invest in financial is intrinsically by assets or physical assets characterized. a information If interpret the investment situation of asymmetry. we decisions of agents in terms of traditional portfolio allocation between insiders theory, the distinction and outsiders introduces a in the The traditional sort of discontinuity portfolio analysis. behalf the assumption that managers act on of generical shareholder discontinuity in the decision because removes this portfolio it us to. regard the single share as a portion of ownership of allows imply loss the firm. But this may a of relevant information, to the level to buy extent that the of wealth needed the control of a firm 188 Assumption ii) and iii) do not imply that the financial control that of a firm cannot be the object of a transaction: they only say the existence of a stock market does not necessarily imply the for financial "outsiders" existence of a market control, since may in regard their investments shares as a financial portfolio. The transactions involving financial control could be modelled as single described episodes of bargaining, or as modifications of coalitions "insiders", but be among a (small) number of they will considered exogenous for the purposes of this analysis. On the basis of assumptions ii) and iii), the asset denominated "control of, the firm" can be defined as: 00 L {u°(k, Iµi) (3 V(0) = exp [-J (D(s)ds] w - A[I(t)]} dz 00 where, like in the model of chapter 5, f(s) is the (instantaneous) cost of firm's financial funds, A[I(t)] is the "purely technological" adjustment cost function of -investment. This function is twice continuously differentiable, with A(0)=0; A'>O; A">O. Equation 3 simply expresses the fact that, by definition, the has the how controlling group right to choose to allocate the flow of variable profits. The value of the asset denominated "control' of the firm X" be the for the firm might not same controlling group of X and the firm, fact, controlling group of another say Y. In since the firm variable profits u°(k, wlpx) of X are conditional on the is high compared to the wealth that an individual might be willing to invest in a diversified portfolio of financial assets. In other for words, in the absence of a market control associated to the the decision investing in stock market, of physical capital is decision equivalent to a of entry. 189 conjectures cx and the market share shx of firm X, it might well be that, because of the strategic market interactions, if the management of firm'Y bought the control of firm x, the increase in the present value of the variable profits of firm Y could be bigger than V(O) such as defined in equation 3. In other words, due to market strategic interactions, the value attributed to the asset denominated "control of firm X" by the management of firm X might not necessarily coincide with the value attributed to the same asset by the management of firm Y. In this sense, a non-hostile takeover might take place when the management of - firm Y offers to the controlling group of firm Xa price for the larger control of firm X that the present value of the future In variable profits. this sense, an element of indeterminacy is introduced for what concerns the value of the firm's financial is it capital. If this true, might not be appropriate to equate the value of the physical capital of the firm to the value of its financial assets, even if one assumes (following the efficient financial markets hypothesis) that the price of the shares of a firm is determined by the present value of the firm's future profitslB. This last point takes into account the fact that serious objections between on the equality stock prices and present value of future been dividends have raised by the empirical studies on excess 18 At this point, an obvious objection is that a situation of different value attributed to the firm's control by the managements firms be of two different would unstable and temporary, since it takeovers would cause mergers and and disappear. However, the firm have incentives managers of a might not to reveal the impact of its Furthermore, a merger on conjectures. some of the advantages of be by a merger could obtained a collusive strategy, which would dilemma. raise a prisoner's 190 by Shiller [1981], [1984] Summers [1986]19. volatility, started and However, the result of the model illustrated later could hold even if it is assumed that share prices are equal to the net present dividends. value of future Assuming that there is no market for control, the problem for the managers is just to raise, in the cheapest way, the financial funds necessary to cover the optimal physical capital, according to the constraint 1. The price in the shares of their company is only that it the relevant to the managers to the extent allows financial funds in the future by issuing possibility of raising new is defined follows: shares. The yield on shares as D(t) Ps(t) (4 r*s(t) = + Ps(t)N(t) Ps(t) If the managers are only concerned with remunerating the the shareholders at expected ex ante rate of return on shares r*s, the higher the rate of growth of the share price, the lower the dividends that the managers need to pay in order to keep the the their the level. If remuneration of shares of company at market the we identify the managers with controlling group, we can regard the decision of profits retention as a redistribution of income internal to the firm, that makes unavailable for the shareholders to the the income not belonging controlling group a portion of by the firm. Such income increase produced a portion of could the firm (and the the firm value of the as a consequence value of be in the future, the basis control) or could reallocated on of a decision taken by the controlling group. In this sense the dividend 19 For an exhaustive analysis of some of the main contributions on [1989]. this regard see Shiller 191 distribution can be thought of as the cost that the management has to support in order to raise external finance on the stock market. For given values of r*s, ps(t), ps(t), N(t), equation 4 yields the dividend policy that the managers must follow in order to the financial remunerate shareholders at market yield. If we assume (on basis a stable "efficient market" relationship the of the information available for investors) between r*s and the risk-free interest rate rf, r*s also represents an opportunity cost for the invested in controlling group's wealth the asset defined as "control of the firm". Therefore, the dividends will be: D(t)= r*s(t)Ps(t)N(t)-Ps(t)N(t) (5 The managers are constrained to choose their dividend policy in the order to remunerate shareholders at the market yield on "market" is shares20. Once such a constraint satisfied, the managers behalf the retain the remaining profits on of controlling group. In this case, if the cash flow is not sufficient to pay the required level of dividends, the firm could pay the shareholders by reducing (reserves), the accumulated profits or the shares. However, on the basis of the balance sheet constraint of equation 1, such a has to be financed reduction of the reserves by issuing new debt, or by the level new shares, or reducing of physical capital. The 20 We can imagine that, if the elements affecting the time path of for length time, in ps(t) act, a sufficient of such a way that the to the capital gain element prevails extent that D(t)= r*s(t)ps(t)N(t)-ps(t)N(t) <0 "negative then, this situation of dividends" would correspond to a firm finds it situation where the convenient to issue new shares on the stock market. 192 reduction in the level of physical capital may bring about the liquidation of the firm. The liquidation of the firm may happen not only in cases of bankruptcy, but also when the controlling group finds the opportunity cost of keeping its wealth endowment invested in physical capital higher than the net present value of the right to dispose of the future flows of variable income of the firm. We define now the cost of the firm's own capital as the negative cash flow that the firm has to pay in order to provide itself with this source of capital. Therefore: c(t) = r*s(t)ps(t)N(t)-ps(t)N(t) (6 where c(t) stands for cost of own capital. in unit terms we will have: c(t) i(t)' = (7 E(t) + R(t) where E(t) and R(t), are respectively the subscription value of the shares and the reserves originating either from the past accumulated profits retentions, or, again, by mean of shareholders' subscription. Therefore, the cost of the firm's own capital could be defined as follows: r*s(t)Ps(t)N(t)-Ps(t)N(t) i(t) (8 t J[(t)_(rf+(Q))_D(t)]dt E(t) + 0 where n(t) are the variable profits, net of "purely technological" investments, adjustment costs-of at time t, D(t) the dividends such in 5, the interest as defined equation rf rate on risk free assets, O (n) a risk premium on borrowing which is assumed to be (as in the 193 increasing function the leverage t2 previous chapter) an of ratio $'>0. The denominator 8 that such that 0(0)=0 and of equation shows the own capital is increased by the past accumulated profits, which, in their turn are determined by the profits, net of the "purely technological" adjustment costs of investments and of the borrowed If the remuneration of capital, as well as own capital. debt payment of the remuneration of and own capital entirely exhaust firm does If the n, then the not accumulates profits. remuneration leads financial flows of debt and own capital to greater than n, then the denominator of 8 increases and the cost of own capital also increases. This mechanism, however, needs further explanation which in 5, dealing the financial decisions will be given section with of the firm. Assumption "v" is analogous to the one of the model of the previous chapter. 3. A digression on the cost of capital and the "financial side of the firm" The considerations contained in this section will not be formalized into a model, but will be helpful in a qualitative discussion that will follow in the next section. Let us consider let that equation 8 and us assume ps-reflects the net present value dividends (assuming financial but of the future efficient markets) the information about it only gradually spreads and affects ps. For a increase in given exogenous value r*s, a persistent the profits less than associated with a complete adjustment of ps(t) to the new level of profits, would enable the firm to retain more profits, 194 reducing more and more the denominator of equation 8. If the average cost of financial funds is an increasing function of the leverage ratio, then the firm's cost of financial capital would be increasingly reduced. if we imagine a persistent reduction (increase) of n°(k, wipi), and we assume that news about this spread first gradually and then increasingly faster (i. e. according to an "epidemic" diffusion model like the one earlier mentioned and described in Shiller [1989], chapter 2), in such a way as to determine increasingly negative (positive) values of ps(t)N(t), this might indeed reproduce (on the basis of equation 8) a few empirical phenomena, such as in levels dividends persistence the of and cheaper cost of own firms capital for that enjoy a persistent positive trend in their level of profits. ; The assumption of persistence of the value s(t)N(t) for a non infinitesimal length of time would also be consistent with the in (Shiller empirical phenomenon of excess volatility share prices [1981], [1984], [1989]) for the same argument that accounts for the "sticky dividends". empirical phenomenon of In fact, looking at if firm is increase equation 8, a able to its level of profits, and this determines an increase in the share price ps, which takes ; that place gradually, so a positive value of the term s(t)N(t) for infinitesimal length persists a non of time, then for the same length of time the firm, in order to remunerate the shareholders at the exogenous rate r*s(t), would need to increase the dividends less than proportionately than the increase in ps. This would allow the firm to retain profits and accumulate reserves, so that, if the cost is function of borrowed capital a of the leverage ratio, even after 195 the complete adjustment of the dividends to the new equilibrium level of the share prices p, the firm would be enabled to enjoy a lower cost of capital. it is clear that the assumption concerning the relation between heavily n(. ) and 'ps relies on the assumptions concerning the diffusion of the information concerning the profits and the profitability of the firm. In addition, the profits and the firms profitability of the could also affect the. risk premium on external financial funds, which might apply not only to debt, but also to shares, to the extent that the presence of a controlling lack for group and the of a market firms' control lead the managers and the other agents to regard the shares as external funds. Therefore, looking at equation 8, one could say that the effect of ) an increase in n(. on the cost of financial capital might be depend ambiguous and on the assumptions on how ps reflects the information on profits, and whether and how the ex ante yield on (being fund shares contains some risk premium an external like the debt). In fact, looking at equation 8, if an increase in profits determines an immediate and instantaneous increase in the share. if the ) prices (and all of profits n(. net of the adjustment costs for investments are entirely absorbed as remuneration for the debtholders and the shareholders), then the numerator of equation 8, 8 and the denominator of would not change. On the other hand, if kind r*s(t) contains some of risk premium negatively correlated to k, the 8 n(. ), or to then numerator of would become smaller, and the denominator larger. All of these possible cases could be summarized kind functional by assuming some of link between n(. ), ps and r*s(t), whose quantitative effects have to be determined iT 196 simultaneously to the firm's financial decision and the flow of funds equation. However, some form of transaction costs might be captured by a "transaction premium" on external funds, not to be confused with the traditional risk premium on borrowing, usually assumed to be an increasing function of the leverage ratio. Such "transaction premium" on external financial funds might be thought of as being negatively correlated with the flow of profits, and formalized as follows: Assumption A. In addition to the risk premium on borrowing $=0(fl), the cost of external financial funds contains a "transaction" premium defined as follows. 02=02(Tt); with 0210; (9 where it are the variable profits. Since the variable profits, as we will see later, are assumed to be a monotonically increasing function of the stock of capital k, Assumption A is equivalent to the following: Assumption B. In addition to the risk premium on borrowing $=0(0), there is a "transaction" premium on firm's external finance defined as follows: Ok=$k(k)' with OkzO; (9' Assumption "A" may be justified on the basis of the following "c" "d". arguments "a", and Assumption B can be justified on the basis of argument "b". a) "Signalling and profit retention" argument: as Leland and Pyle [1977] point out, with information asymmetry and signalling on 197 investment "owned" by insiders stock market, the proportion of the is a signal of good quality of the investment. If the firm has to financing investments by issuing choose between new new shares or decision retaining profits, then the to retain profit to finance new investments and give up dividends may be equivalent to the insiders' decision to keep a high proportion of investment. Therefore, 'the decision concerning profits retention and dividends distribution firm to However, firms in enable the send a signal. operating a context close to perfect competition may not send such a signal, by definition, the firm to the since profits, only allow remunerate financial capital with a return close to the interest rate acting as Therefore, the firms degree opportunity cost. enjoying a certain of (and, high monopoly power as a consequence a margin of profits) are that firms (or firms enabled to send a signal perfectly competitive The immediate with low market power) cannot send. consequence of this argument would be a sort of binary discrimination between the lemon firms able to reduce the premium on their external finance by issuing a costly signal and those unable to do so. However, if we lenders detect the intensity assume that also of the above-mentioned how big the dividends that the insiders signal (i. e. are are giving is their to invested up, and what magnitude compared the capital in it be the enterprise), could assumed that the higher the profit the firm is in rate, the more successful reducing the risk premium borrowing. the interest rate on This situation could be captured by functional link between the assuming a negative risk premium on-debt (and external finance) and the profits. b) "Transactions and information spreading argument": let-us transaction assume that the concerning the liabilities of the firm 198 (i. e. equities and debt) are a major vehicle of information for spreading the quality of the investments of the firms. This is equivalent to the rather orthodox assumption that prices are the main vehicle of information spreading. Therefore, the higher the outstanding stock of financial capital of the firm and the higher the volume of its transactions, the more diffusion of information on the quality of the firm's investments. If the balance sheet constraint 1 holds, the higher the physical capital and the higher the transactions concerning the financial capital, the more diffused is the information on the quality of the firm's investments. This assumption can be formalized following the "general epidemic model" (Bailey [1957], quoted in Shiller [1984] and [1989]). "It is assumed, first, that new carriers of news (as of a disease) are created at a rate equal to an 'infection rate' p times the number of carriers times the number of susceptibles and, second, that carriers cease being carriers at a 'removal rate' c'. (Shiller [1989], p. 15) This model is quoted by Shiller as a possible tool for interpreting phenomena of information-spreading in stock market. It could be extended to the interpretation of the diffusion of information concerning the profitability of the physical capital of the firm "K" (referring to the balance sheet constraint 1) to the extent that "E", "R" and "B" are the objects of transactions, either internal or external to the firm (since one could think of the decision concerning dividends retention as a transaction internal to the firm, involving the controlling group and the outsiders). In particular, let us assume that the "infection rate" ß is constant, the removal rate z depends on the maturity of the financial assets 199 (and, for simplicity, could be assumed to be constant, as a first approximation), the "number of carriers" of information correspond to the individuals who have been involved in the negotiations having for object "S" and "B", the "number of susceptibles" correspond to all of the potential buyers of the firm's assets (i. e., at least potentially, the entire population). Then, on the basis of the balance sheet constraint 1, all these assumptions imply that the "lemon premium" on the firm's external finance depends negatively on "K" . c) "High vs low transaction costs" argument: if, due to information asymmetries, the internally generated financial flow, (which is the object of a transaction internal to the firm) is subject to lower transaction costs than external funds, then a firm enjoying a high market power (and, as a consequence high profits) can potentially rely on a larger source of "low-transaction-cost" financial capital. d) "Empirical" argument: this can obviously be accepted only if one follows methodological approaches where empirical evidence can be regarded as a possible source of information, specially in those cases where the implications of rigourously microfounded theories are clearly "falsified" by empirics or are unable to account for empirical evidence. The main point of this argument can be introduced on the basis of empirical results, such as the famous one provided by Fazzari, Hubbard and Petersen, which can be summarized as follows: "... Indeed, outside the Fortune 500 companies, the overwhelming majority of bond finance has been obtained historically through private placements, usually with life insurances or pension funds. Two features of private placements are significant. First, they are more restrictive than typical bond arrangements, requiring minimum 200 levels of working capital and stockholders' equity and often limiting dividends payments and capital spending. Second, during periods of tight credit, small and medium-sized borrowers are often denied loans in favour of better-quality borrowers, who could also obtain funds from centralized securities markets. Similarly, bank loans and lines of credit, the typical source of finance for smaller industrial firms, restrict operating flexibility and require particular levels for certain financial operating ratios". (Fazzari, Hubbard and Petersen, [1988], p. 153) It must be said, first of all, that in "non-securitized" financial systems like the Italian one, the volume of financial funds traded on stock markets is hardly relevant at all, compared to the intermediated credit. Therefore the considerations of Fazzari, Hubbard and Petersen should apply, in general, to all firms. Since working capital can be regarded as a function of the sales, one can assume, for simplicity, that it is a function of the profits. e) "Strategic use of financial structure" argument: Loosely literature deterrence speaking, the implications of the on entry (for with the strategic use of financial structure example Benoit [1984], Poitervin [1989a] and [1990]) might suggest that higher higher leverage profit margins could be associated with a ratio (and, more generally, to a more costly financial structure). In fact, as suggested by Poitervin [1990], a costly financial structure is a signal that the incumbent with low costs (high profit margins) addresses to the potential entrant. On the basis of this argument, one could think of a causal link 0, among the risk premium the leverage ratio n, and the profits n like the following: ++ 201 This functional link could be thought of as an effect of the persistence of an equilibrium in a strategic financial signalling [1990]. game analogous to the one described in Poitervin Alternatively, if there is common knowledge of this particular use of financial structure for entry deterrence purpose, the banks could be aware that the use of a more costly financial structure is needed to preserve a high market power, and, therefore, high profit margins. For this reason they could allow more profitable firms to have a higher leverage ratio, without charging them with a higher risk premium on borrowing. In other words, one could imagine a tradeoff between leverage ratio and profits like the following +- = ý(n, n) Argument "e" ("strategic use of financial structure") will not be considered in the following part of the paper because it would require a detailed analysis on the characteristics of the is beyond equilibrium in a financial signalling game, which the purpose of this paper. However, the considerations contained in this section suggest that some form of "transaction costs premium" (different from the traditional risk premium depending on the leverage ratio) might be included in the expected ex ante yield on shares, as well as in the interest rate on borrowing. This possibility will be considered in the qualitative analysis that follows. 202 4. The decision problem of the firm Again, the decision problem is assumed to be recursive, i. e., the firm, which in each moment is assumed to maximize the level of 1pi), variable profits u°(k, w chooses the financial structure that minimizes the cost of financial capital. The managers have to remunerate the shareholders at the ex ante expected yield, and have to pay the lenders at the interest rate on borrowing. The controlling group i's maximizing the value of the asset denominated "control of the firm", i. e. co JexP [- JNi) V(0) = V (t)] {u°(k, w - A[I(t)]) dt (10 0 subject to the following constraints (together with the balance sheet constraint 1): k=I- gk, k(0)>O and lim k(t)ý: 0 (11 t->co JE, u°(k, w/P shi, cj) - A[I(t)] - [rf(t) + 0(0(t))]. B(t) + B(t) + D(t) (12 - R(t) - =0 where rf is the interest rate on risk free assets, O(R) a risk borrowing premium on earlier defined, A[(I)] is the function of (purely technological) adjustment costs of investments (such as A'>O, A">O), D(t) A(O)=O, the dividends, B(t) the level of R(t) the borrowing, reserves, S(t) the shares, and a dot over the differentiation variables indicates with respect to time. All of the above conditions have to be considered jointly with the optimal financial structure condition, i. e. an opportune determination of V such that the firm is equating the marginal cost of borrowing to the 203 marginal cost of the own capital "i". In this regard we introduce here a few slight modifications - with respect to the model of the previous chapter - which simplify the calculations and the exposition of this analysis. Let us define the gearing ratio as B(t) P k(t) then, given the balance sheet constraint 1, we have: B(t) 1 -1- (13 k(t) B(t) +1 E(t)+R(t) which can be written as 1+ N1-= h(Q) (14 c+1 where h(n) is a monotonically increasing function of n. Hence, if we define a risk premium on the gearing ratio, instead of the leverage ratio, let us define an equivalent risk premium ratio defined on the leverage as follows: let e(p) be the risk premium defined on p. Then 0(p) = 6(h(p)) _ $(n). We will have in particular 0(0)-0,9'>0, and, as in the model of the previous chapter, $(0)=0, $'>0. We can then define the optimization of the firm's financial structure as follows: . °= min {(1-p)i(t) + [rf(t) + 0(p)] p) (15 )i Equation 15 is the weighted average of the cost of financial funds, including borrowing and own capital. The problem is resolved recursively: in each moment the firm is optimizing the financial structure by choosing the optimal leverage ratio (which, as we will see later, is determined simultaneously 204 with the cost of own capital, given the flow of funds condition 12). The optimized financial structure determines the rate of discount appearing in the intertemporal problem. However, the optimal rate of discount is conditional on the flows of profits, net of the "purely technological" adjustment costs of investments. Therefore, the rate of discount will be a function of both the state variable and the control variable, as we will see in the following section. 5. The financial decisions of the firm. Considering equation 15, and assuming that the second order conditions are satisfied, the first order conditions will be: d-t°/dµ = rf + e(u) + pO'(p) -i=0 (16 Equation 16 says that the firm is equating the marginal cost of borrowing to the marginal cost of the own capital "i". Let us assume that 6(u) is homogeneous of degree 1, such that pO'(p) ° cO(u) (17 where "c" is a constant, then equation 16 becomes: i- rf - 6(p)(1+c) (16' then: p= O-1((i-rf)/(l+c)) (18 assuming that A(. ) is monotonically increasing and invertible, then equation 18 shows that u is a monotonically increasing function of (i-rf), i. e. the difference between the cost of the own capital and the interest rate on a risk-free asset. Since we have n=h(µ), with h(. ) monotonically increasing in p, then n is a monotonically increasing function of the difference (i-rf). We can then define : 205 t2 = h(p) = h(0-1((i-rf)/(l+c))) = b(i-rf) (19 The leverage ratio is an increasing function of the difference between the cost of own capital and the interest rate on risk-free assets because, for a given rf, the higher the cost of the own capital, the higher the incentive for the firm to borrow and increase the leverage ratio. However, equation 19, which derives from the equality, at the margin, between the cost of own capital and the cost of debt, has to be considered together with the fact that the determination of dividends (which depends on a few exogenous variables containing information on the financial markets) contributes to determine the cost of own capital, on the basis of the assumptions made in section 1. Therefore, we have to consider jointly equations 19,8 and 12. The flow of funds condition 12 may be significantly simplified with the following assumptions: u°(k, w/p JE, shi, cj) - A[I(t)] - n(k(t), I(t)IE, shi, cj) (20 s(t) R(t) - - -f(n) (21 Equation 20 is simply a definition, while equation 21 comes from the definition of the leverage ratio n, by observing that an increase in the debts and a reduction in the reserves (accumulated profits) reduces the leverage ratio. Then we can define: [rf(t)+O(n(t))]"B(t) n(k(t), I(t)IE, shi, cj) - - D(t) - -f(Q) (12' Now, considering equation 8, and substituting the flow of funds condition 12', we get 206 r*s(t)Ps(t)N(t)-Ps(t)N(t) (81 i(t) _ t J[_f(cý)]dt E(t) + 0 then i(t) and n(t) are simultaneously determined by the following system: r*s(t)Ps(t)N(t)-Ps(t)N(t) i(t) _ (8' t J[_f()]dt E(t) + 0 b((i-rf)) (19 The system composed by equations 8' and 19 will only be discussed qualitatively, since it is a non linear system of is differential equations. In particular, the system going to give a i(t) It is simultaneous solution for n(t) and conditional on n(t). link a(t), extremely important to point out that the among n(t), and i(t) crucially depends on the functional link (if any), on one hand between n(t) and ps(t), and, on the other hand, between n(t) and functional links be briefly discussed, by r*s(t). All these need to in 3. recalling some of the considerations contained section Before doing that, we nave to maze a ortet aigression oy noting that be in determine 4°(t) function equation 16 may solved order to as a into of p(t). In fact, solving equation 16 for i(t) and substituting (omitting equation 25, we get for simplicity the symbol (t) for notational convenience) 207 v= 6(u) rf + + Ne'(N) - u2e'(), ) _ t°(u) (22. If further we assume that 0"(p) is null or neglectable, and that by definition reminding we have 0 increasing in p (or in n): +++ V°(h(n)) (23. the Let us consider system composed by equations 8' and 19, and let us make the last (very strong) simplifying assumption that the system has a unique solution for i(t) and n(t)21. On the basis of the relations existing among n(t), r*s(t) and ps(t) three main cases can be depicted. Case A: There is no relation between ps(t) and n(t): in other words ps(t) is exogenous. Such a case violates the standard assumption of efficient markets, stating that the share price is the net present value of the future stream of dividends. In fact, according to the efficient markets assumption, if the dividends are correlated with the flow of profits, then the share price ps(t) should adjust to variations of n(t); and, in particular, if the managers acted in the interest of the shareholders, they should pay out as dividends all of the available flow of profits (net of the "purely technological" investments adjustment cost of and of the interest payments on borrowing), determining a correspondent adjustment of the 'expected However, future dividends. since in this part of the analysis it has first been assumed of all that the managers do not necessarily act 21 The assumption of solution unicity is indeed very strong although very usual and commonly accepted in neoclassical and rational expectation literature. 208 in the interest of the shareholders as such, secondly that they regard shares as an external source of finance and merely remunerate them at the expected ex ante yield on shares and thirdly, given that the empirical evidence raises a few doubts about the assumption of market efficiency, then there is no particular obligation to assume a priori that markets are efficients, that share prices reflect the net present value of dividends, and, above all, that dividend payments adjust completely to changes in the flow of profits. In particular, the strong empirical evidence provided by Blanchard et al. [1993] shows that firms tend to accumulate cash windfall without distributing them to shareholders. In this situation, the unexpected cash flows contributes to increase the reserves and reduces the cost of own capital such as defined in equation 8'. Therefore the case where ps(t) and n(t) are independent and uncorrelated is theoretically possible. In this case, we can see from equation 12' (k (t) (t) increase, the leverage that when the profits n ,I }ii) ratio would decrease. The numerator of equation 8' would not change, the denominator would increase, and, therefore, the cost of own capital would be reduced. This would determine a reduction in the leverage ratio, as we can see from equation 19. In this case, the system of equations 8' and 19 would determine a negative functional link between n(t) and C2(t), on the basis of the assumption of independence between n(t) and ps(t). The negative Functional link between n(t) and t2(t) would result in a negative functional link between °(t) and n(t), on the basis of equation 23. A similar situation would result if, on the basis of assumption A of section 3, one assumed that both the remuneration of the own capital and the 209 cost of debt included a "transaction premium" like the one justified on the basis of arguments "a", "c" and "d" of section 3. In this case, a functional link like r*s=r*s(n) would hold. Therefore, an increase in n(t), being ps(t) exogenous, by reducing r*s(t), would reduce the numerator of 8', as well as reducing its denominator because of the accumulation of profits, as already mentioned. Case B: there is no "transaction cost premium" in r*s(t); ps(t) is positively correlated with n(t), perfectly and instantaneously adjusts to n(t), and profits (net of "purely technological adjustment costs of investments") are entirely exhausted in interest and dividends payments. In this case an increase in n(t) would not determine any 8' accumulation of reserves, the denominator of would not change, while the numerator would increase because of the increase in ps(t) determined by the increase in n(t). This would determine an increase in the cost of the own capital, which would determine, in its turn, an increase in the leverage ratio. Therefore in this case we would have a positive correlation between the leverage ratio and the profits. A similar but stronger effect would take place in the case of "excess volatility" of share prices (Shiller [1984], [1989]). In this case, because of the excess volatility assumption, the share prices would overshoot with respect to the expected level in dividends determined by the original increase in n(t). Ceteris paribus, the increase in the numerator of equation 8' would-be larger than in the case of complete exhaustion of profits into 210 dividends, and complete and proportional adjustment of ps(t) to (t) changes in n . The correlation between n(t) and S2(t) would be uncertain and have an ambiguous sign in the following case. Case C: One or more of the following situations apply. C. i) The correlation bei : ween n(t) and ps(t) is weak, so that there is no complete exhaustion of profits into dividends payment and the accumulation of profits is feasible (so that the controlling group is partially enabled to implement a redistribution of profits in their favour, at the expense of the generical shareholder). C. ii) r*s(t) contains a "transaction cost premium" negatively correlated with ir(t), like the one described in Assumption A of section 3, and an increase in n(t) determines at the same time an Increase in ps(t) and a reduction in the "transaction cost premium". C. iii) n(t) is positively correlated with ps(t), but ps(t) only gradually adjusts to changes in rr(t), so that the term ps(t)N(t) at the numerator of 8' persists for a non-infinitesimal length of time. Situation C. i is ambiguous for what concerns the effect on i(t) of an increase in the profits n(t): the numerator of 8' would increase because an increase in n(t) determines an increase in ps(t), but, since the correlation between n(t) and ps(t) is weak, and there is accumulation of profits, the denominator of 8' would also increase, and the resulting effect is ambiguous in the sense that it is not possible to state wether an increase in n(t) would increase or reduce R(t) without any specific assumption on the link between ps and it. Situation C. ii describes a case where an increase in n(t) would at the same time determine an increase in ps(t) and a 211 reduction in r*s(t), so that it is not clear whether the resulting effect would be an increase or a reduction in the numerator of 8'. Therefore, the resulting effect on i(t) would be ambiguous, even if a well-defined functional link between n(t) and ps(t) allowed us to determine the accumulation (if any) of profits and the variation (if any) in Q(t). For what concerns point C. iii, if an initial variation in n(t) determines a variation in ps(t), which takes place only gradually, and the term ps(t)N(t) persists for a relevant length of time, then the final effect of an increase in n(t) will be indeterminate for several reasons: first of all it will depend on the length of time necessary for ps(t) to reach its long run equilibrium. Secondly, the final effect on i(t) will result (as we can see from equation 8') in an increase both of the numerator (because of the increase in ps(t) determined by the increase in n(t)) and in the denominator (due to the fact that the firm will have accumulated reserves as long as the term ps(t)N(t) persists). By summarizing this discussion, we can say'that Case A would result in a relation between n(t), A(t), and, as a consequence °(t) that could be captured and described (at the price of some simplifications) by a functional link of this kind: A(t) = al(n) (24 +-- and 4° = -D°(n) = 1>°(a1(n)) - °(n). (25 Case B, on the other hand would determine a situation that could be described (again with some simplifications) by the following functional link: tl(t) = a2(n) (26 212 +++ and -D° = 4, °(A) = V(a2(n)) = +°(n). (27 where in case A we have fl=a1(n), and in case B we have n=a2(n). In case C, on the contrary, the effect of n(t) on Q(t) would have an uncertain sign, because a variation in n(t) would determine effects of the opposite sign in i(t). However, apart from the very extreme and unlikely case where the opposite effects exactly compensate each other, n(t) should have, in general an effect on n(t), and therefore n(t), in general, will not be exogenous, differently from what has been assumed in the model of the previous chapter. In the following part of the present analysis we will assume that a functional link between n(t) and n(t) exists, and the two cases of R=al(n) and ß=a2(n) will be taken into account in two pieces of qualitative analysis. 6. The "real" decisions of the firm Having found that definition 8' and constraints 12' and 15, referring to the financial decision of the firm, result in a functional link (either n=al(n) or n=a2(n) according to some specific assumptions on the behaviour of the financial markets), we can now turn to the "real" decisions of the firm, which can be described by the maximand 10 subject to the constraints 11. Again the problem can be solved by following the Hamiltonian approach. Co Max V(0)= exp 4, °(n(t))] {u°(k, w/pIE, shi, cj)-A[I(t)]) dt (10 I L- 0 S. t. kI- gk, k(0)>0 and lim k(t)z0 (11 t->c 213 reminding that n(t) = u°(k(t)Iui) - A[I(t)] We define then d4, °/dn = (dV/dai)'(dai/dn); (28 with i=1,2 according to whether the functional link 25 or 27 applies. The Hamiltonian of the problem is the following: H= exp L- °(n(t))tl (u°(") - A(I(t))] + z. [I(t) - g(k(t))] (29 Since the discount factor is a function of it, which, in its turn, is a function of both the state and the control variable, it is a slightly atypical problem of optimal control, which will be solved without working with the current marginal variations. Assuming that the second order conditions are satisfied as well, the first order conditions will be the following: diý 0 SH/öI=O=-e-'P* (n)t'A'-e-V (n)t" '(-A')]. t+z (30 It do hence d4ý° t "nl e-4°(n)t (31 du The transversality conditions are analogous to the ones of the model of the previous chapter, i. e.: lim z(t)zO, [k*(t)-k(t)]z(t) -0. (32 t->°° It is assumed, in what follows, that the transversality conditions are satisfied. Substituting equations 30 and 31 into the derivative with respect to the state variable yields the following: dz 8u d'° Su 1 - e-ý°(n) +r e-ý°(n) 'n+ z'g dt ök L do ök J hence 214 dz r au dý° (33 = e-'D°(n)t I1 I1 -t ni + z"g dt L Sk L do Let us now define: rl dV° e-4,0(n)t -tn '] - m(n, t)' (34 L do Then conditions 31 and 32, together with the constraint 11, yield the following system: z= A"m(n, t) (35 ý- öu z zog + m(tt, t)' (36 k1 kaI- g'k (37 substituting equation 35 into 36 we get: dI 1 [(4)°+g)'A' Su - (38 dt A" ök] I* = G[(8u°/8k)/(V+g)] (38' The system composed by equations 37 and 38 can be represented by the following state spape form. FIGURE 1 I; K 215 In the graphic, SS is the stable saddlepath. The model is similar to the neoclassical model of investment, apart from the fact that 6u/6k is in this case the "marginal profitability" of capital (and not the marginal productivity of capital), containing information on the conjectures. However, a few significant qualitative differences appear if we look at the effects of a perturbation in u(t), given the interactions existing between this variable, the rate of discount, and the leverage ratio. Let us assume that A'>O; A">0; if we also assume that the profit function u(t) is analogous to the one of the model of the previous chapter, and is homogeneous in k- so that an increasing monotone function f(") exists such that 8u/6k a f(u/k) - then a disturbance in u(t) would also imply a disturbance in öu/ök, and the model may be represented according to one of the two following cases. Let us consider first the case where the functional link 25 applies. FIGURE 2 E~ý K An initial unexpected disturbance in 6u/6k would shift away the locus dz/dt=O from the initial equilibrium E to E'. However, to the 216 extent that °(n) is affected by u(t) (since °(n) is a function of n=[u(t)-A(I(t))]), the initial disturbance may also affect the financial variables of the problem (altering the "slope" of dz/dt) and determining the new equlibrium E". As opposed to the standard neoclassical models of investment, an initial disturbance in u°(k, wJpi) (possibly determined by some exogenous change in the conjectures) affects the equilibrium through two channels: the real channel (i. e. öu/ök) and the financial channel (i. e. °). The latter, in this case, amplifies (because of the sign of the functional link 25) the effects of the disturbances in the "real side" of the firm's decisions adding to the former channel an effect of the same sign. On the other hand, if the functional link 27 applies, the effect on -t° of the disturbance in u(t) would have the opposite sign of the "direct" and "real" effect, as we can see in figure 3. In particular, in this case, the effect of an initial disturbance in u(t) through the "financial" channel (i. e. the effect bringing the possible equilibrium from E' to E") could partially or totally offset the "direct" and "real" effect (i. e. the one shifting the equilibrium from E to E'). Figure 3 shows a situation where the "financial" effect only partially offsets the "real" one. I FIGURE 3 -14b -Ab IZIC) "F-- > K 217 7. Conclusions The purpose of this chapter was to present a qualitative framework of analysis for the investment decision of the firm in a "non-securitized" financial system characterized not only by asymmetric information, but also by the absence of a market for firms' financial control. As in the model of the previous chapter, the financial structure and the flow of profits are relevant is describe feed- variables. This qualitative analysis meant to a back mechanism among profits flow, financial structure, cost of capital and investment decision of the firm. In this context, imperfect competition and market power are not ruled out a priori. The case represented in figure 2, based on the functional link 25, is meant, ideally, to find a connection with the analysis by Cowling and Waterson [1976] on the relation between market power and link from profit margins: in addition to the causal going the market by power to the flow of profits (empirically shown Cowling and Waterson [1976]), one could think of another link, going in the opposite sense, via the financial structure and the mechanism of financing investments, which might be potentially barrier-creating. To the extent that a connection exists between profit margins, cost of finance and investments, the empirical links between market power and profits on the one hand, and the link between financial structure and investments on the other, could be thought of as two faces of the same mechanism. The interpretation of the firm's investment decision as described in figure 2, is consistent with the results of those 218 Keynesian macromodels (eg. Greenwald and Stiglitz [1988], [1989], [1990a], [1990b], Greenwald, Stiglitz and Weiss [1984], Bernanke and Blinder [19881), where the financial side of the economy amplifies the possible disturbances affecting the real side of the economy. The case represented in figure 3 shows a situation where the "financial" side of the firm determines a mechanism which tends to smooth out the effects of the perturbations in the "real" investment decision. Both the case of figure 2 and figure 3 would suggest that the flow of profits only is a relevant explanatory variable for the firm's investments, since Q(t) would be a function of the flow of from happens profits, differently what in the model of the previous chapter. This last consideration might also provide a further interpretation for the fact that the empirical analysis performed in the previous chapter does not provide, for the clothing sector, results consistent with the (standard) theoretical framework for be trivial: employed. The reason this could quite simple and the complexity created by some of the potential causal and functional links discussed in this chapter are ignored in the simplified framework of the more standard theory, which might often tend - for the sake of simplicity and algebraic tractability of the models - to focus on a'few relevant causal links. 219 Bibliography of Chapter 6 Abel, A. B., Eberly, J. C., [1993] "A Unified Model of Investment Under Uncertainty", NBER Working Paper Series no. 4296 Aoki, M., Leijohnufvud, A. [1988] " The Stock-Flow Analysis of Investments" in Kohn and Tsiang: "Financial Constraints, Expectations and Macroeconomics", Oxford, Oxford University Press. Bailey, N. T., [1957], "The Mathematical Theory of Epidemics", London, C. Griffin. Benoit, J. -P., [1984], "Financially Constrained Entry in a Game With Incomplete Information" RAND Journal of Economics, 15, pp. 490- 499. Bernanke, B., [1983] "Non-Monetary Effects of the Financial Crisis in the Propagation of the Great Depression", American Economic 73 257-276 Review, vol. pp. . Bernanke, B. e Blinder, A. (1988) "Is it Money or Credit, or Both, or Neither? Credit, Money, and Aggregate Demand" American Economic Review: Papers and Proceedings, may; Princeton University, New Jersey Bernanke, B., Gertler, M. [1989] "Agency Costs, Net Worth, and Business Fluctuations", American Economic Review, vol. 79, pp. 14-31. Bernstein, J., I., Nadiri, M. I., [1986] "Financing and Investment in Plant and Equipment and Research and Development", in Peston, M. H., Quandt, R. E., "Prices, Competition and Equilibrium", Oxford, Phillips Allan Publishers. Bernstein, J. I., Nadiri, M. I. [1993], "Production, Financial Structure and Productivity Growth in U. S. Manufacturing", NBER Working Paper Series No. 4309 Blanchard, O. J., Lopez-de-Silanez, F., Shleifer, A., [1993], "What Do Firms Do With Cash Windfalls? ", NBER Working Paper series no. 4258 Blundell, R., Bond, S., Meghir, C. [1992], "Econometric Models of Company Investments", in Matyas, L. and Sevestre, P. (eds. ) "The Econometrics of Panel Data - Handbook of Theory and Applications", Kluwer Academic Publishers, Dordrecht, Holland. Bond, S., Meghir, C. [1992] "Dynamic Investment Models and the Firm's Financial Policy" no. 92-14, UCL Discussion Papers in Economics, Department of Economics, University College London. Brander, J. A., Lewis, T. R., [1986] "Oligopoly and the Financial Structure - The Limited Liability Effect", American Economic Review, vol. 75, pp. 956-970. Brioschi, F., Buzzacchi, L., Colombo M. G., [1990], "Gruppi di 220 imprese e mercato finanziario", Roma, La Nuova Italia Scientifica. capoglu, G. [1991], "Prices, Profits and Financial Structures A Post-Keynesian Approach to Competition", Hants, U. K., Edward Elgar. Cowling, K. [1982] "Monopoly Capitalism", London, MacMillan Press. Cowling, K., Waterson, M., [1976], "Price-Cost Margins and the Market Structure", Economica, May. Diamond, D., [1984] "Financial Intermediation and Delegated Monitoring", Review of Economic Studies, vol. 51, pp. 393-414. Diamond, D., [1989] "Reputation Acquisition in Debt Markets", Journal of Political Economy, vol. 97 pp. 828-862. Fazzari, S. M., Hubbard, G. R., Petersen, B. C. [1988], "Financing Constraints and Corporate Investment", Brooking Papers on Economic Activity, vol. 1, pp. 141-206. Frasca, F., Marotta, G., [1988], "La Ristrutturazione finanziaria delle grandi imprese", in Atti del seminario "Ristrutturazione economica e finanziaria delle imprese", Roma, Banca d'Italia. Ganoulis, I., [1991], "Stock Market Prices and Investment when Financial Markets are not Strongly Efficient", Applied Economics Discussion Paper Series no. 128, Institute of Economics and Statistics, Oxford University. Gardener, E. P. M., (1991) "Nuove strategie per le banche europee", Banca Impresa, Societä Anno X, 1, aprile; 19-39; Bologna, I1 Mulino, Gardener, E. P. M. e Molyneux, P., (1990) Changes in Western European Banking; Unwim Hyman, London Gertler, M. (1988] "Financial Structure and Aggregate Economic Activity: An Overview", Journal of Money, Credit and Banking, vol. 20, pp. 559-588. Greenwald, B. C., Stiglitz, J. E., [1988] "Imperfect Information, Finance Constraints, and Business Fluctuations", in Kohn and Tsiang: "Financial Constraints, Expectations and Macroeconomics", Oxford, Oxford University Press. Greenwald, B. C., Stiglitz, J. E., [1989] "Financial Market Imperfections and Productivity Growth", NBER Working Paper n. 2945. Greenwald, B. C., Stiglitz, J. E., [1990a] "Asymmetric Information and the New Theory of the Firm: Financial Constraints and Risk Behaviour", NBER Working paper n. 3359. Greenwald, B. C., Stiglitz, J. E., [1990b] "Macroeconomic Models with Equity and Credit Rationing", NBER Working Paper, n. 3533. 221 Greenwald, B. C., Stiglitz, J. E., Weiss, A., [1984] "Informational Imperfections in the Capital Market and Macroeconomic Fluctuations", American Economic Review, vol. 74, pp. 194-199. Hayashi, F. [1982], "Tobin's Marginal q and Average q: A Neoclassical Interpretation", Econometrica, vol. 50, pp. 213-224 Henley, A., [1990], "Wages and Profits in the Capitalist Economy - The Impact of Monopolistic Power on Macroeconomic Performance" Aldershot, U. K., Edward Elgar. Leland, H., Pyle, D., [1977] "Information Asymmetries, Financial Structure and Financial Intermediation", The Journal of Finance, vol. 32, pp. 371-398. Litner, J., [1956], "The Distribution of Incomes of Corporations Among Dividends, Retained Earnings and Taxes", American Economic Review, 46, pp. 97-113. Malecot, J. -F. [1992], "Modeling Companies' Dividend Policy Using Account Panel Data", in Matyas, L. and Svestre, P. (eds. ) "The Econometrics of Panel Data - Handbook of Theory and Applications", Kluwer Academic Publishers, Dordrecht, Holland. Marris, R. [1991], "Reconstructing Keynesian Economics with Imperfect Competition -A Desktop Simulation", London, MacMillan. Mayer, C. (1989) "The Influence of the Financial System on the British Corporate Sector", Mimeo prepared for the conference "The Separation of Industry and Finance and the Specialization of Financial Intermediaries" at Universitä Bocconi, Milano Mayer, C., [1992] "Corporate Finance", mimeo, forthcoming in The New Paigrave Dictionary of Economics and Finance Mayer, C. [1993] "Ownership", mimeo, University of Warwick Myers, S. C. (1984] "The Capital Structure Puzzle", The Journal of Finance, vol. 39, pp. 575-592. Myers, S. C., Majluf, N. S. [1984], "Corporate Financing and Investment Decisions when Firms Have Information that Investors Do Not Have", Journal of Financial Economics, vol. 13, pp. 187- 221. poitervin, M. [1989a], "Financial Signalling and the 'Deep Pocket' Argument", Rand Journal of Economics, 20, pp. 26-40. poitervin, M., [1989b] "Collusion and Banking Structure of a Duopoly", Canadian Journal of Economics, XXII, no. 2, pp. 263- 277. poitervin, M. [1990], "Strategic Financial Signalling "International Journal of Industrial Organization", 8, pp. 499-518. 222 Precious, M. [1987], "Rational Expectations, Non-Market Clearing and Investment Theory", Oxford, Clarendon Press. Ross, S. [1977] "The Determination fo Financial Structure: The Incentive Signalling Approach", Bell Journal of Economics, vol. 8, pp. 23-40. Shiller, R. [1981], "Do Stock Prices Move Too much to Be Justified by Subsequent Changes in Dividends? " American Economic Review, vol. 71, pp. 421-436 Shiller, R. [1984], "Stock Prices and Social Dynamics", Brooking Papers on Economic Activity vol. 2, pp. 457-510. Shiller, R. [1989] "Market Volatility", Cambridge, Massachussets, M. I. T. Press. Summers, L., H. [1986], "Do Market Prices Accurately Reflect Fundamental Values? ", The Journal of Finance, vol. 41, pp. 591- 601. Stiglitz, J. E., [1974] "On The Irrelevance of Corporate Financial Policy", American Economic Review, vol. 64, pp. 851-866. Telser, L. G., "Cutthroat Competition and the Long Purse" Journal of Law and Economics, 20,1966, pp. 26-40. 223 CHAPTER 7 CONCLUDING REMARKS 224 CONCLUDING REMARKS This study is concerned with the macroeconomic effects of market concentration and of institutional factors. Some chapters, namely the second and the third, were meant to show (by means of theoretical and empirical analyses) the effective macroeconomic relevance of the institutional and structural aspects under consideration; other chapters (namely the fourth) were meant to provide evidence in favour of a few starting assumptions of this analysis (in particular, a behavioural assumption of banks' liquidity preference consistent with the "credit view"), or to describe and critically discuss (by means of theoretical and empirical analyses) the interactions between banks and credit suppliers, and their implications for the firms' investment decisions, given the "realistic" hypothesis of simultaneity between investment and financial decisions, and given some "institutional" assumptions typical of the "non-securitized" financial systems concerning the relevance of the securities' transactions compared to the intermediated credit, and whether or not the market for shares is associated to a market for firms' control. In accordance with the methodological premises briefly illustrated in the first chapter, some of the starting points of the analysis are taken from acquired results of the relevant literature. In this sense Stiglitz' [1991] point that economics is a "cumulative discipline", and that "not every piece of research has to begin at the beginning" is taken. The two models presented in the second chapter describe an economy characterized by industrial firms oligopsonistic in the market for credit. In the first (simplified) model, the behaviour of 225 the supply function of bank credit to the industrial firms is captured by a "Cobb-Douglas", which can be thought of as a reduced form for the monetary sector. In this simplified case, an exogenous decrease in the market power of the industrial firms on the credit market increases the effectiveness of monetary policy. This happens because a reduction in the degree of concentration, by reducing the spread between the marginal productivity of capital and the cost of capital, creates an expansionary effect. The second model attempts to generalize the results by introducing a banking sector where banks behave in accordance with the portfolio allocation theory. In this case the results are weakened. In fact, while it is still true that, apart from extreme cases, reductions in the market power of industrial firms in the credit markets increase the macroeconomic level of investment and affect the monetary policy multiplier, on the other hand the sign of the latter effect becomes ambiguous and depends on the analytical forms of the behavioural functions and on the sensitiveness of the agents to the different interest rates of the various assets. Both models however show that modifications of the market structure in the banking sector have, in general, macroeconomic effects. In this sense the second chapter provides a "strong" result. The third chapter suggests an interpretation of the phenomenon of "securitization" on the basis of Williamson's [1985] contractual framework. It is pointed out that in securitized financial systems substitutability between securities and intermediated credit is an empirically relevant phenomenon that makes the demand for bank credit to the industry more unstable than the supply. For this reason, in a securitized financial system, and in a monoequational 226 framework it is possible to identify and estimate a supply function of bank credit to industry, rather than a demand function, while a demand function is identified in a non-securitized system. A comparative econometric analysis has been performed with British and German data, because the two countries had (apart from the phenomenon of securitization) many similarities in their regulatory systems, as well as in the degree of concentration of their banking sectors and in the magnitude of the respective economies, at least until German Unification. In accordance with the predictions of the theoretical analysis, a supply function for bank credit to industry is estimated in the U. K., while a demand function is estimated in Germany, as expected. The diagnostic tests and the statistical properties of the estimated functions are largely satisfactory. The analytical form of the bank credit supply function is based on the theoretical part of the paper by Bernanke and Blinder [1988], which put a strong emphasis on the macroeconomic effects of the attitude of banks and their willingness to lend money to firms, affected by the degree of risk of the whole economy, such as perceived by the banking system. This specific aspect of the behaviour of banks is analyzed in Chapter 4, which contains an empirical analysis (performed with Italian data) of the free liquidity ratio for commercial banks. The free liquid reserves of commercial banks are regarded as a liquid asset associated to the non-investing decision of the bank. Such a non-investing decision by might be determined an increase in the degree of risk of the whole economy perceived by the banks (which is the core of Bernanke [1988] and Blinder's model), and is interpreted on the basis of the recent literature on investment decisions under conditions of 226 framework it is possible to identify and estimate a supply function demand function, of bank credit to industry, rather than a while a demand function is identified in a non-securitized system. A comparative econometric analysis has been performed with British and German data, because the two countries had (apart from the phenomenon of securitization) many similarities in their regulatory systems, as well as in the degree of concentration of their banking sectors and in the magnitude of the respective economies, at least until German Unification. In accordance with the predictions of the theoretical analysis, a supply function for bank credit to industry is estimated in the U. K., while a demand function is estimated in Germany, as expected. The diagnostic tests and the statistical properties of the estimated functions are largely satisfactory. The analytical form of the bank credit supply function is based on the theoretical part of the paper by Bernanke and Blinder [1988], which put a strong emphasis on the macroeconomic effects of the banks attitude of and their willingness to lend money to firms, by degree affected the of risk of the whole economy, such as perceived by the banking system. This specific aspect of the behaviour of banks is analyzed in Chapter 4, which contains an (performed empirical analysis with Italian data) of the free liquidity ratio for commercial banks. The free liquid reserves of commercial banks are regarded as a liquid asset associated to the non-investing decision of the bank. Such a non-investing decision might be determined by an increase in the degree of risk of the whole economy perceived by the banks (which is the core of Bernanke and Blinder's [1988] model), and is interpreted on the basis of the recent literature on investment decisions under conditions of 227 investments' irreversibility and uncertainty (for instance, Dixit [1992a], [1992b], Pindyck [1991]). Even in Chapter 4 the empirical results are consistent with the suggested interpretative framework, and the diagnostic tests, as well as the statistical properties of the estimated functions, are largely satisfactory. Chapters 5 and 6 look at the interactions between industrial firms and financial intermediaries in a "microeconomic" perspective. The focus is on the investment decision, and one of the main concerns is to perform a theoretical and empirical analysis on the connections between risk, cost of capital and investment decisions. Chapter 5 presents a theoretical model - partly based on a contribution by Bernstein and Nadirs [1986] - where the decisions concerning the financial structure and the investment of the firm are simultaneous. Chapter 5 also contains an empirical analysis of firms' investment the decision based on the proposed theoretical The framework. estimates are performed by means of panel data techniques using Italian firms' data referred from three industrial sectors: chemicals, electronics and clothing. Only the results of industries the first two seem to be consistent with the theoretical framework employed, while the inconsistencies of the empirical results referred to the clothing sector seem to be determined by the (possibly complexity of the alternative) causal links among market structure, investment and financing decisions suggested by various contributions in finance as well as in industrial economics. The study of such causal links is precisely the concern of Chapter 6, which contains an analysis of the implications of a few alternative hypotheses (based on precise results of the industrial ýw 228 economics literature) on the link existing between the cost of capital, the market structure and profit margins. Chapters 2,3 and 4 seem to provide some "strong" results (at least as far as the empirical results in the case of chapters 3 and 4 can be regarded as strong, in the light of Stiglitz' [1991] caveats - discussed in the introduction - on the valuation of a theory on the mere basis of "goodness of fit" and other econometric tests). Chapters 5 and 6, on the other hand, raise a few issues that increase the complexity of the models. For this reason the analysis of those two chapters, intended as a" deliberately qualitative critical discussion, might be regarded as a starting point for some further study based on the assumption of simultaneous financing and investment decisions in a non-securitized financial sector. 229 Bibliography of Chapter 7 Bernanke, B. S., Blinder, A. S. [1988], "Is It Money or Credit, or Both, or Neither? - Credit Money and Aggregate Demand", American Economic Review, vol. 78, pp. 435-451. Bernstein, J. I., Nadiri, M. I., [1986] "Financing and Investment in Plant and Equipment and Research and Development", in Peston M. H., Quandt, R. E., "Prices, Competition and Equilibrium", Oxford, Phillips Allan Publishers. Dixit, A., [1992a], "Irreversible Investment with Uncertainty and Scale Economy", mimeo, April, Princeton University. Dixit, A., [1992b], "Investment and Hysteresis", Journal of Economic Perspectives, Winter. M. "Financial Structure Aggregate Economic Gerlter, -[1988], and Activity: An Overview", Journal of Money, Credit and Banking, vol. 20, pp. 559-588. Mayer, C., [1989], "The Influence of the Financial System on the British Corporate Sector", mimeo prepared for the Conference "The Separation of Industry and Finance and the Specialization of Financial Intermediaries" at the Universitä Bocconi, Milano, Italy. Mayer, C. [1992], "Corporate Finance", mimeo, forthcoming in The New Palgrave Dictionary of Economics and Finance Mayer, C., [1993], "Ownership", mimeo, University of Warwick, U. K. Pindyck, R. [1991], "Irreversibility, Uncertainty and Investments", Journal of Economic Literature, September, 29. Stiglitz, J. E., [1991], "Alternative Approaches to Macroeconomics: Methodological Issues and the New Keynesian Economics", NBER Working Paper no. 3580. Williamson, 0., [1985], "The Economic Institutions of Capitalism" New York, The Free Press.